1. Contextualism

One of the most striking developments in recent analytic philosophy is the enormous popularity of the approach to meaning and content known as “contextualism”. Contextualist theories are key players in a range of current debates in philosophy of language, epistemology, moral philosophy, the philosophy of logic, metaphysics, and elsewhere. An orientation that 25 years ago had at most a handful of adherents is now mainstream, in some precincts verging on orthodoxy.

What is contextualism? What it is to have a contextualist view of something? Consider the following three cases:

1) Imagine two acquaintances of a certain Mary, Naomi and Didi, who have independently learned that Mary has won a million dollars in the lottery. Didi is impressed by Mary’s windfall, and says to one of her friends, “Mary is rich.” Elsewhere, Naomi, who moves in more rarefied circles, says to one of her friends, “Mary is not rich at all.” Both Naomi and Didi have probably spoken the truth, for “it is very plausible that the truth of their claims about wealth turns on whatever standards prevail within their conversations” (Richard, 2004, p. 218).

2) Pia has a Japanese maple. She doesn’t cotton to the reddish hue of its leaves, so she paints them green. The task completed, she says, “That’s better. The leaves on my tree are green now.” She has spoken the truth. A few moments later her botanist friend Bill calls, soliciting samples for a study on the chemistry of green leaves. Pia says, “The leaves on my tree are green. You can have them.” Now she has spoken a falsehood (Travis, 1997, p. 89).

3) Suraj and his wife drive by their bank on a Friday to deposit a paycheck. But the lines are long and there’s no particular rush to get the check deposited. Suraj suggests bringing it by the next morning. His wife points out that many banks aren’t open on Saturdays. Suraj says, “I know it’ll be open. I was there two weeks ago on a Saturday.” His avowal of knowledge is true.

Now consider a slight variation. As before, Suraj and his wife drive by their bank on a Friday to deposit a paycheck and are met with long lines. In this scenario, however, it is extremely important for them that the check be deposited before Monday. Suraj makes the same suggestion to wait until Saturday as in the first version and similarly appeals to his previous Saturday visit to the bank. But his wife counters, “Banks do change their hours. Do you know the bank will be open tomorrow?” And Suraj replies, “Well, no. I’d better go in and make sure”. His denial of knowledge is true (lightly adapted from DeRose, 1992, pp. 913-4).

On a traditional conception of truth, the truth value of a claim is determined by two things: the content of the claim, and the way the world is. A claim is true iff what is claimed (content) is in fact so (world). Granting this conception, it’s very difficult to see how to explain Richard’s, Travis’s and DeRose’s proposed assignments of truth values except by supposing that the relevant predicate term in each case—“rich”, “green”, and “knows”—expresses a different content in its two occurrences. Briefly:

a) In 1), the world is held fixed. Given the law of noncontradiction, it follows that Naomi’s and Didi’s claims cannot both be true unless what Naomi claims, in saying “Mary is not rich”, is not the negation of what Didi claims in saying “Mary is rich”. And where could the content differ except in respect of what is expressed by “rich”?

b) In 2), the world is also held fixed. There is a complication in this case, however: Pia’s two claims are made at slightly different times. It is open to hold that her two claims differ in content in virtue of implicitly speaking of different times. (Imagine a silent “now” at the end of the second assertion to parallel the explicit one at the end of the first.) But this enables us to assign a univocal content to the two occurrences of “green” only if we are prepared to say that Pia’s leaves change color, in the relevant dimension, while she walks inside to answer the phone. And that seems ridiculous. The upshot, it would seem, is that Pia’s two remarks cannot differ in truth value unless what she said to be so of her leaves, when she first called them “green”, is not what she said to be so of her leaves when she next called them “green”.

c) In 3), the world is not held fixed; we are imagining two different ‘possible worlds’. But according to a familiar conception of knowledge, whether one knows a given proposition is a matter of what grounds or evidence one’s belief in the proposition is based upon. And Suraj’s grounds for believing the bank to be open do not appear to differ in the two scenarios. Since the two worlds are thus similar in all relevant respects, we are in the same situation as with 1): we cannot explain how both assertions could be true without positing a difference in the content expressed by the two uses of the predicative term.

If we generalize these results in the ways the examples seem to invite, we arrive at three contextualist doctrines, one about talk of wealth, one about talk of color, and one about talk of knowledge. The claim in each case is that the contribution of a certain verb or common noun to the content of an assertion is context dependent: what one asserts of something, in applying that noun or verb to it, will depend upon the context of one’s assertion. Indeed, two assertive uses of the predicative term can different in content to the extent that one such assertion is true and another is false even while the rest of the content is identical and the world is held relevantly fixed.
Some claims of context dependence are uncontroversial. Few people deny, for example, that the content of an assertion of a sentence containing “I” depends upon the context of use; in particular, it depends upon who the speaker is. (Clearly my utterance of “I am cold” might be true while your simultaneous utterance of that sentence be false, and the natural explanation is that my utterance speaks of me and yours of you.) The contextualist doctrines most widely discussed by philosophers these days are not pegged to obvious indexicals like “I”, “here” or “now”. Rather, as in our three examples, they focus on terms where it would be in the nature of a discovery that these terms introduce context dependence into the contents of the assertions they are used to frame. Isn’t it surprising that there is more than one thing we could be saying of a person in saying that she “knows” that such-and-such?

In what follows, I will not attempt to determine the truth of our sample contextualist doctrines, or of any other. Instead, I want to examine a very abstract idea about meaning that might be taken, and has indeed been taken, to provide support for contextualist doctrines across the board. The idea is that use determines meaning. What this idea is taken to suggest is that we can expect context dependence of the sort represented by our three samples to be quite prevalent—-to be, in fact, the characteristic and perhaps even ubiquitous condition of discourse in a natural language. I will give reasons for thinking that the idea does not provide any basis for this expectation.
To see the role the idea can be taken to play in underwriting contextualism, it will help to first look at what epistemological contextualists have had to say about skepticism.

2. Skepticism and doubt

Epistemological contextualism holds that the content of a knowledge ascription, as expressed by an assertion of a sentence of the form “S knows that p”, depends upon the context of the assertion (in ways that go beyond any context-dependence, such as it is, traceable to features of the expressions substituting for “S” and “p”). Details vary, but the most familiar version of the view holds that the content of an assertion of “S knows that p” is such that its truth value hinges in part on whether S is in a position to rule out salient doubts about the truth of the proposition p. Context dependence enters the picture because which doubts count as salient—which “counter-possibilities” are “relevant”, in the original terms of the literature—is held to vary with the context of the assertion.
Much of the appeal of this sort of view for epistemologists lies in its apparent potential to defuse arguments for skepticism. Suppose epistemological contextualism as just described is correct, and suppose also that “it is only in exceptional circumstances that the utterance ‘I might be dreaming’ … can be understood as educing a relevant possibility” (Putnam, 2001, p. 5). It follows from these suppositions that my inability to rule out that I am dreaming would not bear on the truth of what I assert when I utter, in an ordinary, non-exceptional context, “I know that there is a fire going in the study”. Since that doubt is not salient in that context, the truth conditions of my claim to know do not require my being able to rule it out. More generally, everyday knowledge ascriptions cannot be impugned by traditional skeptical arguments. Such arguments purport to deny knowledge precisely on the basis of the subject’s alleged inability to rule out hyperbolic doubts like the dreaming hypothesis, but hyperbolic doubts are not salient doubts in everyday contexts.

How does the contextualist know what I just quoted Putnam as telling us: namely, that the possibility that one might be dreaming is not a relevant possibility, not a live or salient doubt, for our everyday, non-exceptional claims to know things?

The obvious answer is that the contextualist notices that in ordinary life we do not in fact take such possibilities into account when evaluating knowledge claims—and that, moreover, if someone were in the course of everyday life to challenge our claim to know something on the basis of the dreaming hypothesis (or some other skeptical possibility), we would dismiss their challenge as inappropriate and bizarre. Charles Travis (1989, p. 183) writes, “What one says in speaking, on an occasion, of A’s knowledge (or ignorance) that F is determined by what, if anything, does count on that occasion as knowing that F. What so counts is what our reactions show to count.” What our reactions show, for Travis and Putnam and many others, is that skeptical doubts are not, on the great majority of occasions for speaking of people as knowing things, “real doubts” (1989, p. 60) , and so not doubts the putative knower must be able to rule out in order to be correctly said to “know”.

But this answer prompts a further question: do all our “reactions” really show this? Granted, we do not consider skeptical possibilities when offering and assessing knowledge claims in the ordinary run of life. And we would find it at best a weak joke if someone were to bring such a possibility up in the course of an ordinary conversation, in which a knowledge ascription had been proffered with a view to, say, sharing information, planning a strategy, or critiquing someone’s actions. But it is equally true—equally an undeniable empirical fact about human beings and their practices—that many people are impressed and puzzled by skeptical arguments when they are presented with them. They cannot easily accept the skeptical conclusion, of course, but they are struck by the apparent force of the skeptical arguments. And they take those arguments to challenge precisely our ordinary, everyday claims to know things. The contextualist looks at only some of our reactions to a given knowledge claim, those evident ‘in the moment’, as it were, while ignoring others—including our reactions when we worry over our everyday knowledge claims in light of a skeptical challenge. In so doing, the contextualist seems to forget the possibility of second thoughts.

Contextualists might appear to have a response here. They can hold that these second thoughts will inevitably miss their supposed target. When we consider a skeptical argument, we change the context. We make it, in Putnam’s term, “exceptional”. And so if we reach a conclusion about my knowledge that I might express with an utterance of “I didn’t know that there was a fire going in the study after all”, the conclusion I express might very well be true. It might be true because in considering the dreaming hypothesis, we may have made it a live or salient doubt, and thereby rendered the content of a knowledge ascription such that the subject must be in a position to rule out the dreaming hypothesis if the ascription is to be true. But if that is what has happened, then my denial that I “know” is not the negation of my earlier, ordinary claim to “know’, for my earlier claim to “know” did not require for its truth that I be able to rule out the dreaming hypothesis. Thus the truth of my skeptical denial of knowledge does not entail the falsity of my ordinary avowal of knowledge. Granted, it can seem to people that skeptical conclusions reached via skeptical arguments contradict ordinary claims to know. That was certainly how it seemed to Descartes and Hume—that’s why their skeptical arguments troubled them so much. Anyone who has ever supposed that skeptical arguments attack commonsense views about the scope of our knowledge has been under the same impression. But the contextualist can hold that this impression, however widespread, is an illusion. Thus Keith DeRose (2006, p. 321) suggests that it shows only that “speakers are to some extent blind to the context-sensitivity of ‘knows’.”

This maneuver has been much discussed in the literature. Perhaps it provides an adequate response to some of the objections that have been raised to the contextualist dissolution of the skeptical attack on ordinary knowledge. But it does not actually answer the question I raised above. That question was why, given that many people do take skeptical arguments to challenge ordinary claims to know, we should suppose that the “reactions” of producers and consumers of knowledge ascriptions reveal that skeptical doubts are not salient doubts—not “real”, not “live”, not “relevant” doubts—for such ordinary claims, and hence that it is not required that subject be able to rule them out in order to be truly said to know. There is an assumption at work here about which thoughts and activities on the part of human beings constitute the relevant “reactions” to ordinary knowledge claims, and which do not, and my question asks after the basis of that assumption. The line of thought just discussed does not justify or explain the assumption. It simply endeavors to trace out further implications of accepting it.
So where does the assumption come from?

3. Content and use

According to DeRose, the assumption is supported by a plausible view about the relationship between meaning and “ordinary usage” (2005, p. 190).
DeRose asks us to consider the “crazed theory” that “a necessary condition for the truth of ‘S is a physician’ is that S be able to cure any conceivable illness instantaneously” 2005, p. 190). This ‘theory’ is obviously wrong, but, DeRose asks, “in virtue of what is our language in fact such that” it is wrong? He answers the question thusly:

[I]t seems eminently reasonable to suppose that such facts as these, regarding our use, in thought and speech, of the term ‘physician’, are centrally involved: that we take to be physicians many licensed practitioners of medicine who don’t satisfy the demanding requirement alleged; that we seriously describe these people as being ‘physicians’; that we don’t deny that these people are ‘physicians’; that claims to the effect that these people are ‘physicians’ intuitively strike us as true; etc. It’s no doubt largely in virtue of such facts as these that the traditional view, rather than the bizarre conjectures we are considering, is true of our language: the correctness of the traditional view largely consists in such facts (2005, p. 291).

We are justified in taking these intuitions, reactions and uses to reveal the truth conditional content of utterances and sentences because those intuitions, reactions and uses in large measure constitute our utterances and sentences having the truth conditional content that they do.
In particular, “such facts about ordinary usage … provide us with our primary, most important and best evidence” for the context-sensitivity of given terms because those facts “are also that in which the context-sensitivity of those terms consists” (DeRose, 2005, p. 291). Applying the point to the predicate that interests him the most, DeRose concludes:
‘Knows’ is context-sensitive … largely because speakers in some contexts do (in fact, with propriety, and with apparent truth) seriously describe subjects as ‘knowing’ propositions when those subjects meet certain moderate epistemic standards with respect to the propositions in question, even if they don’t meet still higher epistemic standards, but, in other contexts, will go so far as to (in fact, with propriety, and with apparent truth) seriously deny that such subjects ‘know’ such things, reserving the ascription of ‘knowledge’ only for subjects that meet some more demanding epistemic standard’. (DeRose, 2005, p. 291)

Here, then, is a genuine answer to the question I raised in the last section. The reason we should not treat our reactions to skeptical arguments as telling us something about which doubts are live or salient for claims to “know” in everyday contexts is this: such reactions do not play a role in determining the truth-evaluable content of our everyday-context claims to “know”. The reactions that do play a role in determining the content are the actual applications of the term we make and accept in those everyday contexts. The different pattern of application we evince in the context of a discussion of a skeptical argument, by contrast, determines the content of those applications. To put it bluntly, the thought is that our tendency, in some contexts, to seriously describe people as “knowing” propositions when they meet certain conditions makes it the case that it is true, in those contexts, to say that they “know” those propositions. And our reluctance, in other contexts, to describe them that way makes it the case that, in those contexts, it is false to say that they “know”.

DeRose represents these views as instances of the principle that ordinary usage determines meaning, and that principle certainly seems difficult to deny. The meanings of expressions in a natural language are not determined by intrinsic features of the sounds and shapes of these expressions, nor are they assigned by divine command. They can come only from how we speakers of the language in fact use, and have used, these expressions. Whatever Joseph Smith or the characters in the Cratylus may have thought, this principle is not open for debate. But it is crucial to see how special is DeRose’s vision of what a meaning-determining “fact of ordinary usage” will look like. His “facts of ordinary usage”, with respect to a given expression F, are facts about which objects we apply F to in a given context and which objects we do not. In effect, they are facts, relative to each context in which we might have occasion to apply F, about how we are disposed to apply F in that context. The principle that use determines meaning is thus taken to yield the result that the truth-evaluable content of our assertions will align with our immediate dispositions in the use of the terms involved. And if our dispositions vary with context, then so will content.

I will (with obvious prejudice) call the view that content is determined by the sorts of facts DeRose mentions the superficial use-theory of meaning. As a general picture of how content must accrue to assertions, the theory is incorrect. The alignment it insists upon between content and disposition need not obtain. A range of intelligible and plausible ideas about the character of the content of given areas of discourse predicts divergences between content and disposition, and suggests in any case that simple dispositions to apply or refrain from applying needn’t be the primary basis by which content is determined.

I will discuss one model for such a divergence. It is provided by the account of natural kind terms introduced by Putnam (1975a, b). Putnam pointed out that members of the extension of a natural-kind term F needn’t fit the ‘stereotype’ associated with that term—i.e., the constellation of observable characteristics an ordinary speaker of the language associates with F. Membership in the extension thus cannot be a function of conformity to this stereotype. What members of the extension share, according to Putnam, is rather an underlying nature, which can be fully described and explained only in the vocabulary of some natural science—but which needn’t actually have been thus described and explained by anyone at the time that F is in currency. These features are impossible to square with the view that F’s extension is determined by ordinary speakers’ intuitions and dispositions concerning the acceptance and rejection of assertions of the form, “S is F”. Such intuitions and dispositions may reflect only the stereotype we associate with F, which in turn may correspond only loosely with the term’s actual extension.

Putnam’s account is perfectly compatible with the principle that use determines meaning. There may be a temptation to suppose otherwise, based on a thought to the effect that any extension-determining meaning we can ‘give’ to an expression, through our uses, reactions, dispositions, or intentions, must be such that we are in a position to adjudicate membership in that extension. But this proto-verificationist thought at best reflects a failure of imagination. So long as we possess, at least implicitly, the concept of a natural kind, and so long as we are in a position to draw attention to exemplars of a given natural kind (for example, demonstratively), we are perfectly capable of ‘giving’ an expression a meaning such that it stands for that natural kind construed on Putnamian lines. Indeed, there is nothing to prevent us from doing so explicitly and intentionally: consider “Let ‘F’ stand for that [natural] kind of thing” said while pointing at some instances thereof, a definition that certainly might be misunderstood, but then again, might not.

At the same time, Putnam’s account is frequently characterized as portraying the meaning of a natural-kind term as ‘world-involving’ or ‘extension-involving’, and these descriptions are apt. For on this account, if we set about to answer in any meaningful detail a DeRose-type question about “in virtue of what our language is such that” a natural-kind term has the particular extension that it does, we will quickly find our investigation shifting from an analysis of the relevant bits of language to consideration of the natural kind itself. What facts about speakers of English—about their usages, intentions, reactions and so forth—determine, on the current conception, is no more or less than this: that the English word “lemon” speaks of lemons. If, knowing this fact, we still have residual questions about the constitution of the extension of “lemon”—e.g., about why this object (which is green) should fall into that extension but that object (which is an etrog) should not—then these are questions about what it is to be a lemon, which is to say, questions for plant biologists, not for field observers of ordinary language use.

These points, at a suitable level of abstraction, generalize beyond the case of natural-kind terms. Consider DeRose’s example of “physician”. In our culture at the present time, there is an extremely complicated set of social institutions surrounding the practice of medicine. There are institutions involved in training physicians, in providing them with support, space and equipment, in overseeing their treatment of patients, in managing payment for their work, in orchestrating their collaborative endeavors, and so on. Inextricably entwined with all of this are rules and procedures for credentialing physicians. Physician-hood is about as thoroughly institutionalized a status as it is nowadays possible to achieve. Anyone who claims to be a “physician” while lacking the ordinary licenses, degrees and affiliations had better have some very special reasons for that claim if it is to have a chance at being true. The same goes for anyone who would deny that a person with the requisite licenses, degrees and affiliations is a physician. Ordinary speakers can be expected to differ in the extent to which they are familiar with the ins and outs of the credentialing and related procedures, and hence, for their intuitions and dispositions concerning the application of the word “physician” in particular cases to correspond to the relevant facts about the practice. And even if their intuitions and dispositions do line up neatly with the relevant institutional facts, it will not be the intuitions and dispositions, but rather the institutional facts with which they align, that explain why “physician”, being a word for physicians, has the extension that it does.

Again, there may be an inclination to protest that our words have only the meanings that we give to them, and it must therefore be ‘up to us’ whether a given object falls into the extension of a given word. And no doubt, there is some sense in which this claim is true. But in whatever sense it is true, it cannot conflict with the evident fact that we are often interested in kinds in the world around us whose natures we imperfectly understand, and hence whose boundaries we can individuate at best imperfectly. Given our interest in these kinds, it is natural that we should have words for speaking of them. We must then in some sense or other ‘give’ words meanings suitable for so speaking of them, and intuitions and dispositions on our part, as well as descriptive conditions in our possession, will certainly play a role in constituting their possession of such meanings. But in giving words meanings suitable for speaking of such kinds, we allow facts about the kinds themselves to determine appropriate extensions. The superficial use-theory seems plausible only if we reject this possibility—I think, necessity—and hence only given a too thin vision of the ways in which our words can engage the world of which they speak.

What about “know”? I do not claim that “know” is a natural-kind word like “lemon” or a social-institutional one like “physician”. But these examples have motivated an abstract point: our words can fix on properties and kinds that have a life at least partly independent of how we are disposed to apply those words to the objects we encounter. That is one way we can use words, and so one sort of meaning use might determine. “Know” might have such a meaning. Until we have been given a reason to think it does not, we have been given no reason to suppose that “facts about ordinary usage” of “know”, as DeRose conceives and constrains that category, have overriding authority in deciding upon the content “know” expresses. We have been given no reason to refuse to treat the results of skeptical reflection as “reactions” on our part to our ordinary knowledge claims. But without a basis for that refusal, the contextualist attempt to deflect the skeptical challenge, as outlined in the previous section, is unpersuasive. The skeptical challenge must fail, we can all agree. But we need a more satisfying understanding of why.
A more general moral is that we should be suspicious of the widespread tendency of contextualists of all stripes to move from observations about how we are disposed to apply terms directly to conclusions about the contents expressed by those terms. If there is a justification for this mode of argument, it is not to be found in the bare idea that use determines meaning.

4. Conclusion

I close with two tentative further thoughts.

First, the superficial-use theory of meaning makes it easier for us to be right in the things we say. It makes it easier for us to be right in the things we say because it entails that, in effect, we make ourselves right. In having the dispositions of application that we do, we shape the contents of our utterances in such a way as to go some distance toward ensuring their correctness. That might seem a pleasing prospect. The problem is that the easier it is for us to be right in this way, the less there is for us to be wrong about. And that means the less there is for us to find in the world to talk and think about and possibly come to understand. Only if our utterances are aptly interpretable as attempting to place the objects and people we discuss in categories constituted independently of the dispositions of use informing our immediate discursive context will those utterances be beholden to a subject matter whose objectivity promises no end of challenges for our understanding.

Second, Wittgenstein is famously associated with the idea that meaning is determined by use. It is no surprise that many contextualists claim him as their predecessor. But Wittgenstein is no superficial use-theorist. His conception of use is tied to the idea of institution and practice, and crucially, institutions and practices do not reduce to the applications and dispositions of a moment. For Wittgenstein, the context that matters, in understanding what we do with language, is broad and open-ended. The idea that the content of a given assertion will automatically be determined by dispositions specific to that local discursive context alone, and so that contextual variation in dispositions of application automatically bespeaks contextual-dependence of content, is foreign to him. I close with two passages, one familiar, one less so, that make this point well:

33. … What, in a complicated surrounding, we call “following a rule” we should certainly not call that if it stood in isolation.
34. Language, I should like to say, relates to a way of living.
In order to describe the phenomenon of language, one must describe a practice, not something that happens once, no matter of what kind.
It is very hard to realize this. (Wittgenstein, 1978, pp. 335-6)

How could human behaviour be described? Surely only by showing the actions of a variety of humans, as they are all mixed up together. Not what one man is doing now, but the whole hurly-burly, is the background against which we see an action, and it determines our judgment, our concepts, and our reactions. (Wittgenstein, 1980, p. 213)

References cited
DeRose, Keith. 1992. “Contextualism and Knowledge Attributions.” Philosophy and Phenomenological Research 52:913-929.

DeRose, Keith. 1995. “Solving the Skeptical Problem.” Philosophical Review 104:1-52.

DeRose, Keith. 2005. “The Ordinary Language Basis for Contextualism, and the New Invariantism.” The Philosophical Quarterly 55:172-198.

DeRose, Keith. 2006. “‘Bamboozled by Our Own Words’: Semantic Blindness and Some Arguments against Contextualism.” Philosophy LXXIII:316-338.

Hawthorne, John. 2004. Knowledge and Lotteries. Oxford: Oxford University Press.

Putnam, Hilary. 1975a. Is Semantics Possible? in Mind, Language and Reality: Philosophical Papers, volume 2 Putnam (1975c). Originally published in 1970.

Putnam, Hilary. 1975b. The Meaning of ‘Meaning’. in Mind, Language and Reality: Philosophical Papers, volume 2 Putnam (1975c).

Putnam, Hilary. 1975c. Mind, Language and Reality: Philosophical Papers, volume 2. Cambridge: Cambridge University Press.

Putnam, Hilary. 2001. “Skepticism, Stroud and the Contextuality of Knowledge.” Philosophical Explorations 4:2-16.

Richard, Mark. 2004. “Contextualism and Relativism.” Philosophical Studies 119:215-242.

Travis, Charles. 1989. The Uses of Sense: Wittgenstein’s Philosophy of Language. Oxford University Press.

Travis, Charles. 1997. Pragmatics. In A Companion to the Philosophy of Language, ed. Bob Hale and Crispin Wright. Oxford: Blackwell pp. 87-107.

Wilson, Mark. 2006. Wandering Significance: An Essay on Conceptual Behavior. Oxford University Press.

Wittgenstein, Ludwig. 1978. Remarks on the Foundations of Mathematics. revised ed. MIT Press.

Wittgenstein, Ludwig. 1980. Remarks on the Philosophy of Psychology. Vol. II University of Chicago Press.

It is thought to be a platitude that “one can be morally responsible only for one’s own actions.” Call this Platitude. For many philosophers, Platitude entails what I call Slogan: moral responsibility presupposes personal identity.[1] In other words, for someone now to be morally responsible for some past action Φ, this person must be the same person as the agent of Φ. Those philosophers who accept Slogan disagree over specifics. Some believe that what responsibility presupposes is numerical identity, while others believe that what it presupposes is narrative identity. And in the former group, there is even more disagreement about which specific criterion of personal identity properly applies to responsibility. What they all do agree on, though, is that responsibility presupposes identity of some kind, that some version of Slogan is true. Nevertheless, Slogan is false. Moral responsibility has nothing to do with personal identity, of any stripe. This is the case even granting the truth of Platitude. It is my aim in this paper to show why.

Background Issues

What is meant, first of all, by “moral responsibility”? There are multiple conceptions.[2] Because most adherents of Slogan are primarily interested in discussing the metaphysics of personal identity, though, which conception they have in mind typically remains underspecified. A clear understanding is essential, though. What most often comes out is a worry about accountability, and this is sometimes taken to mean the following: I cannot appropriately be held to account, that is, punished, for some action (read: “crime”) unless I was the person who performed it.[3] This way of putting it might suggest that worries about reasons of justice or fairness are at the root of Slogan: I cannot justly be punished for someone else’s crimes. This cannot be right, though. For one thing, accountability isn’t restricted to criminal activity. Not only do we seem to hold each other to account for bad actions that are not crimes (for example, laughing at a “friend” behind his back), but we also hold one another accountable for positive actions, actions for which we express our gratitude or admiration, say. For another thing, once we admit this last point, we can see that, while considerations of justness or fairness are surely relevant to our expressions of blame—it would indeed be unjust to sanction someone wrongly, for example, for something no one in fact did—they do not seem to constitute our most fundamental normative considerations in assessments of accountability. To see what I mean, consider the fact that it wouldn’t necessarily be unjust, say, to praise someone wrongly for something, for example, for me to express gratitude to you for what I thought was your pushing me out of the way of an oncoming car when in fact that was done by a strong gust of wind. But while my praising you isn’t unjust or unfair, it is nevertheless inappropriate in another sense. The inappropriateness consists in its not being fitting, that is, you don’t merit praise. So it must be the felicity conditions for one’s being praised or blamed for Φ that include one’s being identical to the agent of Φ. And if any of the felicity conditions for being held accountable for Φ are not met, that would be sufficient for one’s not being accountable for Φ.[4]

What, then, are these felicity conditions? I take Platitude to state one such condition for being accountable for Φ, namely, Φ must be attributable to the accountable agent: the current accountable agent must be its “owner.”[5] It would not be fitting, in other words, for me to hold you accountable for some past action if it weren’t your own, that is, if it weren’t properly attributable to you qua agent.[6] Otherwise, holding accountable would be senseless. This is because, in the standard case, agents—robust entities still in existence—are held accountable for actions—often merely momentary events no longer in existence—so for the latter to ground an assessment of the former, they must be inexorably tied to the former as their expression in the world.[7] In other words, we do not merely condemn bad actions, say; rather, we condemn bad agents for bad actions, and it makes sense to do so only insofar as these actions are visible, worldly manifestations of those agents in some essential respect. Platitude thus ties accountability to attributability.

Slogan, then, ties attributability to identity. The idea seems natural: if some past action is now mine for purposes of my being held accountable for it, then its “mineness” must consist in my being identical to the agent who performed it. This way of putting it thus calls for an explication of ownership/attributability in terms of a criterion of identity, so believers in Slogan typically attempt to insert the central feature of their favorite criterion of personal identity directly into Platitude, asserting in so doing that what makes an action one’s own is just whatever personal identity consists in. After all, if some past action is now mine in virtue of the fact that I was the person who performed it, then what makes that past person’s actions now mine is just whatever made him me. On this methodology, the plausibility of criteria of identity increases or decreases in tandem with the plausibility of their entailed criteria of attributability. As it turns out, though, and as I will show, all proposed criteria of attributability are implausible. This suggests something deeply flawed about the methodology, a flaw based on a mistaken assumption on which the entire enterprise is founded. In pointing out this flaw, I hope to clear the path for a very different approach to understanding the nature of Platitude and the felicity conditions for accountability thereby.

It is worth testing the patience of some readers for the sake of clarity for all, so let me briefly restate the ostensible relation between the various conceptions here. In Platitude, attributability is taken to be a necessary condition for accountability. In Slogan, the central feature of a criterion of personal identity—its set of necessary and sufficient conditions—is taken to be precisely the central feature of the criterion of attributability, so determining what makes an action one’s own is just a matter of figuring out the right criterion of identity and applying its necessary and sufficient conditions in addition to action ownership. I am here granting the truth of Platitude. The theorists earlier cited take Platitude to entail Slogan. I deny the entailment. To see why, we will allow for now the methodology of these theorists and proceed by attempting to figure out which antecedent criterion of personal identity, if any, best enables us to fill in the following blank: what makes some past action one’s own now is __________________.

Lockean Consciousness[8]

John Locke was the first to draw an explicit connection between accountability and personal identity. He called “person” a forensic term, “appropriating actions and their merit” (Locke 1694). In other words, a person (as opposed to the organism-referencing term “man”) is the only entity to which actions—and praise and blame for such actions—are properly attributable. Because Locke (1694) defined “person” as essentially a self-conscious being—an entity capable of being conscious of its own consciousness—he took the identity of persons across time to consist in a preservation of that same consciousness across time, and so “as far as this consciousness can be extended backwards to any past action or thought, so far reaches the identity of that person.” Insofar as “consciousness” is typically taken to refer to memory,[9] the Lockean criterion of identity is thus, roughly, the following: X at t1 is the same person as Y at t2 if and only if Y’s consciousness extends backwards to X, that is, Y remembers the thoughts and actions of X.

In applying the central feature of this criterion of identity to attributability, Locke filled in the blank as follows: what makes some past action one’s own now is that one now has consciousness—memory—of it. As he puts it, “[C]onsciousness, as far as ever it can be extended … unites existences and actions, very remote in time into the same person, as well as it does the existences and actions of the immediately preceding moment: so that whatever has the consciousness of present and past actions, is the same person to whom they both belong” (Locke 1694). Furthermore, “This personality extends itself beyond present existence to what is past, only by consciousness, whereby it becomes concerned and accountable, owns and imputes to itself past actions, just upon the same ground and for the same reason that it does the present” (Locke 1694). Consciousness thus provides the criterion of attributability in virtue of its provision of the criterion of identity: my consciousness of some past action makes me identical to its agent and so is precisely what renders that action my own, imputable to me for purposes of accountability. Indeed, it is just its purported plausibility in accounting for attributability-with-respect-to-accountability that renders the consciousness/memory relation plausible in the first place as a criterion of personal identity.

In assessing this view (and the others to come), we will be asking the following question: does the proposed criterion of identity in fact deliver the right criterion of attributability-with-respect-to-accountability? If not, and if the plausibility of the proposed criterion of identity depends (at least in part) on its plausibility in accounting for attributability, then the criterion of identity itself loses plausibility.

On this test, the Lockean view fails pretty miserably, as it is susceptible to some very persuasive counterexamples. Consider Drunken Mel: suppose Mel has gotten very drunk and then, when pulled over by the cops, goes on an anti-Semitic tirade against the arresting officer. When he wakes up in jail the next morning, Mel truly doesn’t remember a thing. Is the tirade the night before still attributable to him for purposes of moral assessment? Most people would say yes without hesitation. Locke, however, explicitly says no: if he genuinely didn’t remember saying what he said, then he’s not accountable for it, precisely because it wasn’t his action. But this assessment is deeply counterintuitive. Indeed, there is a further, more troubling point to be made here. As it turns out, some criminal actions may actually cause memory loss. In particular, up to 45% of murderers black out after the crime and wind up losing all memories of it, and this is true of the perpetrators of other violent crimes as well, such as rape and aggravated assault.[10] Clearly, though, we don’t want to let these criminals off the hook. These are still their actions, we want to say, regardless of whether or not they remember them. Consciousness is thus not necessary for ownership.[11]

It is also not sufficient. Consider Criminal Carl: suppose Carl steals a car and then has his memory trace of that action copied into Innocent Ida’s brain. Ida then wakes up from the surgery and seems to remember furtively looking around an unfamiliar street one night, smashing open a nearby car’s window with a tire iron, hotwiring it, and then driving away. This action is clearly not “hers” for purposes of moral assessment, even though she now has consciousness of it.[12]

The central feature of Locke’s proposed criterion of identity—consciousness—cannot constitute the central feature of the criterion of ownership for Platitude. It would seem, then, that this inadequacy renders his criterion of identity less plausible, and so what we need is just a better criterion of identity to deliver the content of Slogan.

The Biological Criterion

According to the Biological Criterion of personal identity, X, a person, at t1 is the same individual as Y at some other time if and only if X is biologically continuous with Y.[13] In other words, as long as some humans X and Y have the same essential animal nature, they are one and the same metaphysical object. Inserting this central feature into the criterion of ownership, then, we get the following: what makes some past action one’s own now is that it was performed by an individual human organism with whom one is now biologically continuous.

This view seems to account for both counterexamples to Locke. In response to Drunken Mel, we want to say that the drunken tirade is attributable to sober-Mel, even though he doesn’t remember it, and appeal to the Biological Criterion could explain this by pointing out that it’s his action in virtue of its having been performed by the same human organism he is now. The view could also account for Criminal Carl, for even if Ida has consciousness of Carl’s thievery, it would still not be her action insofar as she is not biologically continuous with Carl.

Nevertheless, the account is quite flawed. Consider Cerebrum Transfer: suppose that I have robbed a bank, and then my entire cerebrum is transplanted into another body’s brain (itself devoid of a cerebrum). Stipulate that the person who wakes up will be exactly like me psychologically, that he will have apparent memories of the crime, that he will share all of my criminal values, and that he will have inherited my intention to go on a spending spree after waking up (and does so). On our application of the Biological Criterion, he will not be me—my human organism will just not be his—and so consequently my actions will not be his either: it would thus be a mistake to attribute the crime I committed to him. But this conclusion will strike most of us as itself a mistake.[14] Instead, there is still a significant relation between him and me—a psychological relation—that seems clearly to make it appropriate to attribute my actions to him (and thus to hold him accountable for what I have done). To say that my actions are not his will be a great joke to him, believing as he will that he has gotten away scot free with the crime. Biological continuity is thus not necessary for attributability.

Now an advocate of the Biological Criterion might respond, as David DeGrazia (2005) in fact does, by saying that “in the world as we know it,” biological continuity is in fact necessary for the sort of psychological relation on which we seem to be relying. But this reply misses the point. What we are looking for is a criterion of the ownership of actions, an explanatory story about what makes various past actions properly attributable to one’s current self. Biological continuity, even if one necessary ingredient in real-life cases, fails to provide any relevant explanation of what is obviously a psychological matter, though: attributability, and thus accountability, obviously has to do with one’s performance of actions—the execution of certain intentions—as well as one’s being a legitimate subject of various assessments for them, that is, one’s being in a position to hear, be receptive to, and respond to praise or blame, along with one’s being subject to experiencing various emotions associated with these reactions and judging the fairness or appropriateness of the original assessments. But all of these are psychological states, and they are the only states that are relevant to a criterion of attributability.

Biological continuity is also not sufficient for attributability. Consider Sudden Fugue: suppose Johann abruptly enters into a fugue state that lasts two years, during which time he is called “Sebastian.” Sebastian is very different from Johann psychologically, and he spends his time moving from place to place and committing minor immoralities. After two years, Johann abruptly wakes up again without a clue as to where he is or where he’s been. He eventually finds his family but never remembers what happened during the missing two years, despite seeing pictures of Sebastian from various surveillance cameras. If we insist on applying the Biological Criterion to Slogan, the Sebastian-actions would have to be attributable to Johann, but this seems false. A biological approach to ownership just won’t do.

The Psychological Criterion

The natural move at this point, then, would be to embrace and apply the Psychological Criterion of personal identity, according to which X, a person, at t1 is the same individual as Y at some other time if and only if Y is uniquely psychologically continuous with X, where psychological continuity consists in an overlapping chain of a sufficient number of direct psychological connections (which includes memory connections but also appeals to other psychological connections: intentions fulfilled in action, persisting beliefs/desires/goals, and similarity of character traits).[15] Inserting this account into the criterion of ownership, then, we get the following: what makes some past action one’s own now is its having been performed by someone with whom one is now uniquely psychologically continuous.

Even if we restrict our attention to cases “in the world as we know it,” the Psychological Criterion is much more directly relevant to attributability, and thus to accountability, than the Biological Criterion, for it includes those features we think essential to responsibility-assessments, namely, intentions, beliefs, emotional dispositions, and so on. And it is because the Psychological Criterion is also constituted by these other relations, in addition to the memory relation, that it is able to account for all our previously troublesome cases. For instance, it allows us to maintain in Drunken Mel—even more plausibly than in the application of the Biological Criterion—that despite the fact that sober-Mel remembers nothing of his tirade, those are still his actions in virtue of the fact that he is psychologically continuous with drunken-Mel: the two stages remain strongly psychologically connected in all the psychological relations other than memory (compare, Glannon 1998). Furthermore, with respect to Criminal Carl, the application of the Psychological Criterion helps us see why Locke’s view gives the wrong answer: it takes much more than a single memory of an action for it to be one’s own; instead, psychological continuity with that agent requires a significant number of overlapping psychological connections. Alternatively, were there a significant number of those connections actually in place, the psychological continuity established would thereby be sufficient to attach ownership of actions to the psychological continuant, so this view can also explain ownership in Cerebrum Transfer. Finally, in Sudden Fugue, it looks as if there just is no psychological continuity between Johann and Sebastian, and so there would be, on the psychological continuity approach, no ownership relations between Johann and the Sebastian-actions, which also seems correct.

Notwithstanding these successes, though, the view still faces worrisome counterexamples. Consider Gradual Fugue: suppose that the transformations from Johann to Sebastian and back occur gradually (the product of a tumor that grows and then wanes, say), such that psychological continuity is preserved all along. Regardless of such continuity, the Sebastian-actions still don’t seem to belong to Johann: while he now remembers them, he nevertheless repudiates them as the actions of someone with a very different psychological make-up and motivational set, actions that seem to Johann as those of a stranger, inexplicable in motivation and justification. Nevertheless, a psychological continuity view of ownership implies that the Sebastian-actions do belong to Johann. Psychological continuity is thus insufficient for attributability.

It is also unnecessary. Consider Disproportionately Disciplined Doug: suppose that Doug gently mocks a young girl and then is beaten up so badly by her parents that the beaten-up inheritor of Doug’s body in the morning is psychologically discontinuous with Doug (given that there is an insufficient number of psychological connections between him and Doug), but nevertheless suppose the beaten-up person retains a memory of mocking the child and still has a small number of beliefs, desires, and character traits that were nevertheless essential to his being moved to mock her. Here I suspect we would agree that the mocking-action is still attributable to the assault survivor, despite the fact that he is not psychologically continuous with Doug (and so is not Doug, on this view).

What this suggests is that only a certain subset of psychological relations is relevant to ownership. The puzzle, then, is to figure out how to identify that subset. One popular way to do so is with reference to the agent’s attitudes toward certain psychological relations.

The Narrative Criterion

The Gradual Fugue and Disproportionately Disciplined Doug cases, so difficult for an application of the Psychological Criterion to handle, seem tailor made for showing off the strengths of what is known as the narrative identity view. On this account, the type of personal identity relevant to practical concerns like moral responsibility isn’t a matter of our answering a reidentification question—“What makes X at t1 numerically identical to Y at t2?”—but is rather a matter of our answering a characterization question—“What makes various experiences and actions occurring at different times those of one person?” That is to say, on the characterization conception the question of identity and the question of attributability amount to precisely the same question, and the sense of identity involved in the answer given will be the one to which we refer when talking about an identity crisis, an uncertainty about who one really is. To understand the unified personhood delivered by an answer to the characterization question, consider by way of contrast a living, thinking, persisting human organism, something to which various experiences merely happen over time. The difference between such an individual and a full-fledged person is that the various experiences in a person’s life are somehow going to be unified in virtue of them all being attributable to her. To find a criterion of attributability, then, is just to find a criterion of (characterization) identity.

What is that criterion? According to the narrative identity theorist, what makes various experiences properly attributable to one person is that they are “part of a single identity-constituting narrative” (Schechtman 1996). On this view, the unity of narrative persons (or narrative egos) is provided by the agent’s subjective attitude towards—his tale-telling incorporation of—his actions. Consequently, this account of identity just amounts to the following criterion of ownership: what makes some past action one’s own now is its being correctly included as an event in the self-told story of one’s life.[16] As Schechtman (1996) puts it, “What it means for an action to be part of someone’s narrative is for it to flow naturally from the rest of her life story—to be an intelligible result of her beliefs, values, desires, and experiences,” and “[t]he more an action seems to stem from a coherent and stable pattern of values, desires, goals, and character traits, the more it seems under a person’s control [and so determines the degree of accountability assigned].” Alternatively, the more inexplicable some action—the harder it is to fit within the narrative of one’s life—the less accountability one has for it, insofar as it is difficult to see it as one’s own, as flowing from one’s genuine self. Another way some narrative theorists put this is in terms of identification: to the extent one actively identifies with some past action, that is also the extent to which one owns it and is thus eligible for accountability for it.

How does this view resolve the previously puzzling cases? In Gradual Fugue, Johann would not be taken to own the Sebastian-actions, given how inexplicable those actions were within the self-told narrative of his life as a whole. They didn’t at all flow from the beliefs, values, and desires he now has (or had prior to the tumor), and so are in no way representative of him, of the genuine self that Johann had developed (and has since regained). He does not, and likely cannot, identify with Sebastian. In Disproportionately Disciplined Doug, on the other hand, insofar as Doug’s mocking-action flowed from the motivational beliefs, values, and/or desires that persist in the beaten-up person—and he still remembers mocking as well—that beaten-up person may still identify with that action; it thus still belongs to him. In either case, then, the degree of psychological connectedness itself that obtains is far less important than the person’s attitudes toward those various connections.

Narrative identity seems to handle all the other cases as well. In Drunken Mel, insofar as the drunken action flowed from his persisting beliefs and values, Mel could easily incorporate it into the narrative of his life, regardless of whether or not he remembers performing it (as long as he was informed about it—see below), so he would come to own that action, rendering it part of his unified person-life and rendering him eligible for accountability for it thereby. In Criminal Carl, the one action of Carl’s that Ida “remembers” is so incongruent with her own set of beliefs and values that it couldn’t intelligibly be included in her life story, a result again compatible with our intuition that she doesn’t own Carl’s action. Finally, in Cerebrum Transfer, the inheritor of my cerebrum will be exactly similar to me psychologically, and so of course he will inherit the beliefs, values, and desires that were central to my action. Indeed, insofar as he will continue to think he’s me, it will be no surprise when he incorporates my actions into his life’s narrative, rendering them, on this theory, his own.

There are problems with the theory on its own terms (not the least of which is its vagueness), but the main problem for our purposes is that it goes too far in the subjective direction. Suppose Johann comes to identify with the Sebastian-actions after all by incorporating them into his biography (perhaps to explain his current charitable contributions to tumor research). Just because he claims to take on ownership of the Sebastian-actions, however, this doesn’t yet seem sufficient to render them his: we are likely inclined to continue judging them as being just as much not attributable to him as we were when he was alienated from them. Subjective incorporation into one’s narrative is insufficient for ownership.

It is also unnecessary. In constructing my narrative, I may forget several details, edit some out as unimportant, or deceive myself about their happening to me in the first place. Do those actions then cease to be attributable to me? Of course not. Suppose Mel were never informed about his tirade. His failure to remember it, and thus his failure to incorporate that action into his life story, would nevertheless be irrelevant to the fact that he still owns that action.

The mistake being made here is found in an assumption linked to the narrative identity theorist’s sole emphasis on the subjective, first-personal nature of the narrativity at issue, namely, that the process by which various characteristics become properly attributable to me is both active and reflective. In other words, the narrative view says that various past actions are mine only insofar as I gather them up via my active storytelling about my life. But this is a serious mischaracterization of the ways in which our selves are often actually constructed, for much of who we are is determined passively, a process during which we also lack self-knowledge.

Identification, the process by which some psychological trait, experience, or action is rendered one’s own, has long been thought of as distinctively active, a product of my reflective endorsement.[17] But it’s easy enough to think of cases where this isn’t the way it works at all, cases in which all that reflection does is reveal to the agent what he’s already identified with and in which active endorsement is more like a command of the British queen to do what her parliament has already enacted: redundant and ineffectual.[18] To see this, return to Drunken Mel. Suppose once Mel sobers up he both fails to remember the anti-Semitic tirade and he actively rejects ownership of his actions. Even after he views the Cop-Cam recording of the incident, he continues to reject ownership, but at that point he nevertheless starts to notice little things about himself, such as the fact that he has, over the years, avoided friendships and working relationships with Jews who had given him repeated opportunities to do so, that he had a few anti-Semitic jokes in his “repertoire” that he didn’t think were that offensive, and that he often had negative thoughts about Jews when he would see certain names in print. He eventually comes to recognize, then, that he does identify with that anti-Semitic guy on the Cop-Cam, indeed that he had all along, and that that was precisely the action he should’ve expected he would perform when his inhibitions were lowered with alcohol. This recognition, though, doesn’t involve an active endorsement at all; instead, it’s just an admission of a pre-existing identification, one that in fact he had been actively trying to deny, to no avail. It is his action whether he likes it or not, in other words, whether it fits intelligibly into his overall narrative or not.

Now a narrative theorist might well respond that we have shown precisely how Mel’s drunken tirade in fact does fit into an intelligible narrative, that it makes sense only within the framework of an even much longer story, likely including his experiences as a child, listening to his deeply anti-Semitic father denying the Holocaust, say. This response misses the point, however, for just because one might be able to provide an intelligible narrative within which various experiences and actions are united, that’s not at all to say that it’s the narrative itself that unites them, that makes those experiences and actions one’s own, events with which one identifies. Instead, as Mel’s case makes clear, any narrative told in this sort of instance will merely be a way of pointing out preexisting patterns, a recounting and categorization of identifications, not a creation of them. This sort of case, then, is one of passive identification, identification that takes place without one’s knowledge or active consent, and as such it is an objective matter, a fact independent of one’s active endorsement, self-conception, or storytelling abilities.

Of course active identification may well still occur. My point here is simply that narrative identity is irrelevant for attributability, and thus accountability. It incorporates both too much and too little: too much, insofar as there may be actions that are part of a narrative that aren’t properly attributable to the narrator; too little, insofar as it ignores cases of passive identification, and if it does incorporate them it does so only by overlaying a narrative structure onto a pre-existing set of identifications with the actions in question. But in any event narrative identity has nothing to do with what makes various actions my own. Thus, because the narrative identity theorists have explicitly made their theory of identity stand or fall with the plausibility of their criterion of attributability, it falls.

Two Universally Troublesome Cases

We have been exploring whether any theory of identity contains as its central feature some relation that could plausibly serve to explain attributability in a way that grounds the move from Platitude to Slogan. I have argued that none of the theories of identity—psychological, biological, or narrative—does so. On their admitted methodology, then, these theories of identity must be rendered less plausible.

Perhaps, though, there are responses to the counterexamples I have raised, so that one or more of the theories can still deliver Slogan. In this section I want to offer a couple of cases that are going to be difficult for any such Sloganeer to handle, and in the next section I want to diagnose the fundamental problem with Slogan itself. The cases I will discuss here are familiar, but they are targeted at theories of numerical identity, so I need to say something briefly about why narrative identity won’t just escape the net. While narrative identity indeed isn’t a theory of numerical identity itself, nevertheless it in fact presupposes some such theory. After all, in order to persist as narrating persons, we ourselves (the metaphysical units constituting or giving rise to the narrators) must continue to exist (DeGrazia 2005). Consequently, whatever theory of numerical identity it presupposes (it is actually compatible with both psychological and biological views) will have serious trouble with the following two cases.

Start with Physical Fission: suppose that I split down the middle and my two halves re-grow their missing halves, resulting in two people who are exactly similar to me, both psychologically and biologically.[19] Given that the numerical identity relation is one-one, and my relation to the fission products is one-many, the relation between them and me cannot be numerical identity. So it looks as if the only plausible account of this case from within any standard view of numerical identity is that I do not survive, that the two fission products are two new individuals. Adopting any of the criteria of personal identity for the criterion of ownership, then, yields the conclusion that neither fission product owns any of my actions (and if narrative identity presupposes numerical identity, none of the post-fission actions or experiences could be narratively unified with my pre-fission actions or experiences). So if I had performed some immorality and then undergone fission, that pre-fission action could not be properly attributable to anyone anymore, despite the fact that two people would “remember” doing my deeds and relish the feeling of having gotten away with them (and they would, in addition, be exactly like me in every respect). And if attributability is necessary for accountability, neither fission product could be accountable (to any degree) for my actions. This will strike many of us as the wrong answer, however—surely my actions are (at least partially) attributable to both fission products—and so the fission case seems to thwart the viability of any relationship between accountability and numerical identity.

One way to deal with Physical Fission could be to maintain that there were actually two people all along. This is the four-dimensionalist reply (see, for example, Lewis 1976, Noonan 1989, and Sider 2001). Very roughly, the view is that persons have not only spatial parts but also temporal parts. This allows there to be overlapping temporal stages of two (or more) distinct persons. Just as two different roads may overlap for certain spatial stretches, so two different persons might spatially overlap for certain temporal stretches. Indeed, this is the purported resolution of the fission puzzle: my two fission products simply overlapped pre-fission. Both persons were in existence all along, but their temporal stages spatially coincided up until the time they were separated at fission. In this way, then, we can preserve the alleged relation between accountability and identity while also preserving our intuitions about attributability in the fission case: an action performed by the shared person-stage pre-fission is properly attributable to both fission products insofar as both performed it. Indeed, claims Sider (2001), the success of the four-dimensionalist view in accounting for this puzzle case (while preserving all the relevant intuitions and platitudes) provides key support for the view.

In reply, it is important to note first of all that four-dimensionalism is a general ontological picture and not a criterion of numerical identity. Indeed, on four-dimensionalism, the question of what makes some individual at t1 numerically identical to some individual at t2 is simply the wrong sort of question to ask. A person (or human individual, say) is typically thought of as a four-dimensional space-time worm, an object numerically identical only with itself, and any such individual to which we may point at any specific moment in time is technically merely a time-slice, or person-stage, of that general four-dimensional object. The question of identity across time, then, becomes a question about unity-relations for the four-dimensionalist: what makes a person-stage at t1 part of the same person as a person-stage at t2, that is, what unifies those temporally-distinct stages into those of one person?[20]

What’s relevant for our investigation is that the four-dimensionalist ontological view implies nothing whatsoever about the nature of that unity relation (compare, Sider 2001). It could be a biological relation, a memory relation, a more robust psychological relation, or a narrative relation; none is favored by the ontology itself. But what this means is that if one appeals to four-dimensionalism to handle the fission case en route to establishing a relation between identity and accountability, that ontology alone fails to provide any explication whatsoever of what identity (or unity, in this case) consists in. Instead, one must bring to the table one or the other criteria of identity/unity in order to supplement it with the four-dimensionalist ontology. But then the other counterexamples already discussed are still relevant: someone who wants to insert a Lockean, Biological, or Psychological Criterion into the criterion for ownership still has to find a way to deal with those cases. In other words, because four-dimensionalism is silent about the nature of identity/unity, it is silent about the nature of ownership or its relation to accountability. All it does is (perhaps) help one get out of the fission jam, but this isn’t going to be sufficient to help the theories of identity we have discussed get out of their other attributability-related jams. Consequently, while some advocates of four-dimensionalism have touted its facility in dealing with the fission-and-accountability case as a boost for the prospects of the ontology, it actually seems that because of their prior need for a plausible unity relation explicating the relation between identity and attributability (granting that they accept Slogan), these advocates have jumped the gun.

Nevertheless, suppose an advocate of one of our theories of identity were able to construct a unity relation adequate for dealing with all the counterexamples. Were one then to couple that theory of identity/unity with four-dimensionalism, it seems one might well be able to deal with fission, and one could have defended Slogan in so doing. This hypothetical strategy is undermined, however, by consideration of Branch-Line (a variation on Parfit 1984 and Olson 1997). Suppose that I perform some immoral action today and then while I sleep tonight I am secretly duplicated by some mad scientists who then kill me in a few days, cremate my body, and secretly insert the duplicate into my place. The duplicate, it seems clear, cannot be me.[21] Nevertheless, it seems that my actions are still attributable to him for purposes of accountability. The pull of this intuition has its most compelling source in a vivid representation of the phenomenology of the duplicate. He will come into existence full-blown thinking he is me, and he will (quasi-)remember my actions, will delight in thinking that he’s gotten away with the immorality, will carry out my intention to celebrate the immorality, and so forth. Once we imaginatively project ourselves into his mental terrain and appreciate that he is psychologically exactly similar to the real me, it becomes extraordinarily difficult to resist the verdict that my actions are properly attributable to him as well, that he is just as eligible for being held accountable as I am.

Here our intuitions about attributability are essentially the same as they are in the initial fission case, but now a four-dimensionalist rejoinder will lead one astray. This is because identity simply doesn’t obtain between the duplicate and me, so any effort to hitch one’s identity wagon to four-dimensionalism in order to explain how my actions are attributable to the duplicate will wind up with a very implausible criterion of identity/unity (a criterion Olson 1997, p. 60, calls incoherent). No plausible criterion of identity, therefore—even if it can avoid the earlier counterexamples and is supplemented (as a unity relation) by a four-dimensionalist ontology—provides an adequate account of how accountability presupposes personal identity.

Responsibility Without Identity

What explains the repeated failure of attempts to link identity with accountability? I believe it is the fact that, while accountability does presuppose attributability, attributability simply does not entail identity. In this section I will offer two general arguments for this point. The first is that there is a key structural difference between the attributability and identity relations. The second is that attributability may obtain in real-life cases of non-identity.

Start with the structural point. One of the essential general requirements of numerical identity is uniqueness. Numerical identity is a one-one relation, obtaining between one and only one individual at two different times. Ownership/attributability, on the other hand, may be one-many. There are several ways to see this point.

I begin (perhaps rather weakly) with an analogy. Note that the ownership relation in other contexts does not entail identity for precisely this reason. The point is most obvious in the case of property ownership. Consider this platitude: I can be taxed only on my own property, and never on someone else’s. In one sense this is true, but in another sense it is false. It is true that I cannot be taxed on your property if it is exclusively yours; it is false that I cannot be taxed on your property, however, if you and I share ownership of the property in question. We could accept, then, that a person can only be taxed on her own property, and that one person can’t be taxed on anyone else’s property, without having to accept that taxation presupposes exclusive property ownership: identity must be one-one, while property ownership may be one-many, so property ownership doesn’t entail identity.

Similarly, then, we could accept that a person can only be accountable for her own actions, and that one person can’t be accountable for anyone else’s actions, without having to accept that accountability presupposes numerical identity. The reason is that there are two senses in which to take the phrase “I can’t be accountable for someone else’s actions.” It is true that I can’t be accountable for actions that are exclusively yours; it may be false, though, that I can’t be accountable for actions of which we share ownership. Indeed, this seems to be the most plausible description of what occurs in Fission: if I commit an immorality and then undergo fission, both of my fission products would own my actions, even though neither would be me. Identity is a one-one relation, while action ownership, like property ownership, may be one-many, so action ownership—attributability—doesn’t entail identity.[22]^, [23]

Perhaps Fission is too controversial to do the heavy lifting here, though. Turn, then, to everyday examples of non-exclusive action ownership. These are what we call joint actions, for example, singing a duet, having sex, building a house, suing as a class, performing a play, (gang-) robbing a bank, engaging in or winning a tug of war (or any of a variety of team actions).[24] While each of these presupposes individual contributions, the end result, the action, is irreducibly joint. These are actions only a we performs—“Remember when we sang that duet?”—and such actions render us joint agents in the performing. Accordingly, such actions are attributable to the joint agent, that is, the team, the troupe, the gang, the class. Nevertheless, as individual members of these collectives, we each share in the ownership of the actions of the collective, and so may be assessed thereby.

But then here is the key point: it will not be true of me that if I share in the ownership of some joint action, there must be some past agent who performed that (joint) action who is numerically identical to me, for the agent of that past action was a we, not an I.[25] If ownership of such actions entailed numerical identity with the original agent, then by transitivity all the members of the joint agency would have to be identical with one another. Ownership of actions thus does not entail personal identity, for ownership isn’t a uniqueness relation—a single piece of property may have multiple owners, as may a single joint action. But if this is true, Slogan is without warrant: Platitude simply doesn’t entail that responsibility presupposes identity. All that we need to talk about is the relation between attributability and accountability directly. Talk of identity just gets in the way.

Perhaps our doubts about Slogan could have been reached much more quickly and easily, however. This would be the case if one thinks Platitude itself is false, so that one isn’t in fact accountable only for one’s own actions. A pertinent possible example comes from the case of parents, who we might think are accountable for the actions of their children.[26] If parents are indeed accountable in such circumstances, then neither identity nor attributability is relevant to accountability.

This is too quick, however. The advocate of Platitude need merely point out that in discussing such cases, we need to be careful to determine the precise target of the relevant assessments. So suppose you have brought your young children to my house, and while they are running around making a ruckus, they knock over and break one of my favorite Lladro porcelain figures. The young children aren’t accountable for this action. Are you? To the extent that I find you blameworthy (says the advocate of Platitude), it is actually for a different action, namely, your failure to rein in your children, or, more generally, your lax parenting. But in such cases, then, you are accountable for your own actions, that is, the actions properly attributable to you. Of course, you may be morally obligated to replace the figurine or pay me for the damage, but that’s a point about compensation, not responsibility: you may indeed be required to compensate someone for the destructive actions of someone else.[27] But even if I have a right to compensation from you for the actions of your children, that’s not yet to say that you are blameworthy for those actions. The advocate of Platitude thus has a ready reply to such cases.

Nevertheless, there are relevant cases in the neighborhood to consider, and this leads to the second general argument against Slogan. Consider Evil Parent: suppose an evil parent secretly told his children to run around and then break my Lladro. Now this parent is blameworthy. But again, this case doesn’t necessarily cast doubt on Platitude, for the breaking of the Lladro now actually seems properly attributable to the parent. This point may become more obvious when we consider Ordering General: suppose a general commands his troops to “take the bridge.” Despite the variety of individual actions performed by the general’s soldiers that together wholly constitute that action, the taking of the bridge is properly attributable to the general: to the extent he is praiseworthy or blameworthy for that action, it is his.[28] Consequently, even if one does bear an exclusive ownership relation to some action, one doesn’t necessarily have to be involved in the actual performance of the action at all. Here attributability obtains despite one’s complete non-identity with the agent.

What really undercuts Slogan is the fact that accountability merely presupposes attributability, and attributability is just a relation between an accountable agent and an action. Identity, by contrast, is a relation between a person and a person (or a person and an individual). These are, then, just different relations about different sorts of entities, so it’s no wonder that identity and attributability go their separate ways. A criterion of identity is just about the wrong kind of relation and the wrong kind of entities to be relevant to moral responsibility.

Responsibility With Identification

Identity is not necessary for accountability, but we are assuming, along with Platitude, that attributability is.[29] What, then, does attributability consist in, if not the central features of personal identity? This is just to ask the characterization question again, albeit without the problematic narrative identity as its answer. In this final section, I will briefly and sketchily explore an alternative answer to this question, one that builds on the work in the first part of the paper, as I take it to be a condition of any plausible theory of attributability that it account for all the cases we have discussed.

Accountability presupposes the attribution of an action to an agent, something which renders the agent open to moral appraisal (praise or blame) for the action.[30] What is an agent, though, at least with respect to accountability? It might seem as if agents are necessarily performers of actions, but this is doubtful. There could be, after all, continually frustrated agents, those who simply aren’t able (for whatever reason) to execute their wills in action. What does seem essential to agency instead, however, is at least the capacity to have a will to act in the first place.

Now as I said early on, insofar as an accountable agent—a robust entity—is susceptible to being praised or blamed for some action—typically a momentary event—that action must implicate her agency, must somehow be an expression of who she is. It must be, in other words, an action with which she is identified: the agent herself must, in some crucial sense, be found in the action. What, then, does such identification consist in?

Here is where we butt up against a mountain of contemporary literature. The search for the conditions of identification has become, to some extent, like the search for the Holy Grail: many have taken on the quest, numerous seekers have gotten horribly lost in the pursuit, and it remains entirely unclear that the object of desire even exists.[31] I will certainly not try to grasp the cup myself here; rather, I merely want to articulate a few conditions on any future searches. In other words, I want to lay out some of the formal conditions for carving out the nature of identification as it pertains to accountability, given what we have learned to this point.

First, because of the relation between action and agent presupposed by accountability, for an agent to be identified with an action for purposes of accountability (for it to be attributable to her), the action must bear some dependence relation to the essential component of her agency, namely, her will. This is not to say, however, that the action in question must be performed by the agent on whose will it is dependent. The children in Evil Parent and the soldiers in Ordering General perform actions that depend—at least in part—on the wills of the parent and the general respectively: it was in virtue of their willing it that the various actions were performed. In both cases, the children and the soldiers execute the wills of those to whom their own wills are subordinate. Of course, whether or not they insert their own wills into this process (and endorse or reject the subsequent actions) makes a difference for whether or not they themselves identify with (and so own) the actions, but it’s perfectly possible that their endorsement (say) of their leaders’ wills could also leave the original dependence relation intact, such that their actions are overdetermined (by both their own and their leaders’ wills) and their leaders’ ownership of and accountability for those actions is thus not reduced one whit.[32]

Wills aren’t freestanding mechanisms, however. Rather, they operate and are embedded within a much wider volitional web, a web of beliefs, judgments, desires, attitudes, and cares. Indeed, these are the sorts of psychological elements necessary to accountable agency. A computer, after all, could be said to have a will it executes, but that is insufficient for the kind of agency we have in mind. This is because, in being accountable, the agent is open to praise or blame, but it is not the agent’s will that is the appropriate object of these attitudes; rather, they target a complex psychological creature, an entity capable of understanding the point of the appraisal, reflecting on and evaluating its past deeds and its current character in light of the appraisal, and responding (often emotionally) to the demands implicit in the appraisal. We appraise a self, in other words.

For the self to be sensibly implicated in such appraisals, though, it must now be implicated in the will on which the relevant action depends. Take the self’s volitional web to refer to those psychological features that shape, incline, and perhaps even determine its will. The aspects of the self relevant to identification will thus be found in that web. Indeed, the motivating task of many of those dubbed “Real Self View” theorists (see Wolf 1990) has been precisely to identify “the” privileged subset of psychic elements within this volitional web that are truly representative of the agent’s “real” self. Various contenders have arisen, including higher-order desires (desires about what one’s will is to be) (for example, Frankfurt 1971), evaluative judgments (judgments about actions worth willing) (for example, Watson 1975), and cares (for example, D. Shoemaker 2003). Which one is correct is, again, not my concern here. Instead, I am simply adopting the more neutral stance that it is within one’s volitional web that the real self is to be found, and so when that real self is expressed in one’s will, one is identified with that will. Consequently, when some action is dependent on that will, one is identified with that action thereby, and it is this relation between an agent’s real self and an action that makes the action attributable to the agent.

Nevertheless, this accounts only for attributability in the moment of action (where the agent at the time of action has a direct identifying relationship to that action). But judgments of attributability are much more typically made at a later time: some past action is attributable to the agent now. And this fact, I think, is what makes drawing the link to personal identity for the Sloganeers most tempting. Indeed, how can we make sense of it without reintroducing personal identity? Suppose Evil Parent occurred a week ago but I just now discovered the truth. I now hold the parent accountable for those week-ago actions of his children. Given what was said above, though, this would suggest I believe those actions are now attributable to him because (a) those actions depended on a will with which some agent identified last week, and (b) the agent now before me is identical to that week-ago agent.

Unsurprisingly, I disagree. Notice, first of all, that if any identity is involved here, it is just not personal identity of any plausible stripe. To see why, consider Fissioning Faye: suppose Faye assaults a child and then undergoes fission, and because of a botched procedure, both fission products wind up psychologically discontinuous from Faye while still retaining the central aspects of Faye’s volitional web that both motivated the original assault and allow the fission products to replay it over and over again with a tingle of excitement. Neither fission product could be identical to Faye on either the Biological or the Psychological Criterion (and insofar as narrative identity presupposes one of these criteria, its possibility adds nothing here). Nevertheless, both, it seems, are accountable for the assault in virtue of their ongoing ownership of her actions. But then personal identity cannot be doing any of the relevant work.

What might be more plausible, then, is to suggest that agential identity is necessary for accountability. In other words, perhaps there is an identity relation for agents, rather than persons, that unites the evil parent now with the week-ago entity on whose will the children’s action depended, rendering current accountability-assessments of him coherent. The problem with this move is that it plays fast and loose with the term “identity.” We have already seen that the attributability relation necessary for accountability may adhere to neither the uniqueness nor the transitivity requirements of an identity relation, so at best the label would be highly misleading.

However, it is clear that there must be some relation between the current accountable agent and the past agent on whose will the action depended. What precisely that relation consists in, though, is less than clear. I will here merely suggest a plausible possibility, put in terms of Evil Parent: given that the accountability of the week-ago agent for the actions of his children presupposes that he identified with their actions then (that is, they depended on his will), that agent’s accountability may be transferred to the current agent in virtue of the current agent’s identification with that week-ago agent.

This introduces a third target of identification. I have already discussed (1) identification with actions and (2) identification with wills. Now I am talking about (3) identification with agents. The three are intimately related insofar as they all contribute to an explanation of attributability, and thus accountability, with respect to past actions. But they each have different formal conditions. I have suggested that identification with actions centrally requires a dependence relation running from action to will. I have suggested that identification with wills centrally requires the expression of an agent’s self via some shaping or dependence relation between a privileged subset of elements in the agent’s volitional web (the real self) and the agent’s will. I now want to suggest that identification between a present agent and some past agent centrally requires a similarity relation between the privileged subsets of the current and past agents’ volitional webs, a similarity relation that obtains in virtue of a causal dependence relation running from the present agent’s real self to the past agent’s real self.

There are a few elements here to explain. First, a current agent cannot be said to identify directly with a past action, given that that action couldn’t have depended on his will—there is no backwards causation. What the original agent’s identification with his action consisted in, though, was not just the action’s dependence on his will but also his will’s dependence on his real self. But insofar as the current agent may identify (or be identified) with that past agent’s real self, he could be said to have adopted or inherited the actions dependent on that original self as his own, providing the necessary condition for him to be accountable for them thereby. What I am suggesting is that this last sort of identification occurs in virtue of the (presumed) similarity between the current agent’s real self and last week’s real self, where the current agent’s real self is the way it is because last week’s agent’s real self was the way it was.[33]

This general approach accounts for all of our previous cases. It tracks similarity of psychological states and so properly allows for attributability in cases where biological continuity is absent. It also properly allows for attributability in cases where psychological continuity is absent, for it focuses on the similarity only of a certain privileged subset of psychological states whose continuity alone may be insufficient to preserve full-blown psychological continuity. And it properly allows for attributability where narrative identity is absent, for its focus is on similarity between the features of agents’ volitional webs that are in fact relevant to motivation and expressive of the agents’ real selves, regardless of those agents’ subjective attitudes toward those features.

Is this account just another way of smuggling identity back into the picture, though, perhaps by offering an alternative account of identity itself? In other words, perhaps the intuitions buttressing my proposal make it plausible only insofar as they are really about identity themselves, albeit now about identity grounded in my sort of identification, where it involves the persistence of volitional webs or their privileged subsets. I want to make absolutely clear one last time that this cannot be what’s going on here, for two simple reasons. First, the numerical identity relation must obtain uniquely (be one-one), whereas the diachronic aspect of attributability (the third sense of identification) need not. Multiple persons may identify with some past agent’s real self, as in the fission case. Second, the identity relation must be transitive, whereas attributability of the sort I have been discussing is not. I may identify with some past agent’s real self, and that past agent may identify with some other more distantly past agent’s real self, but it may not be the case that I identify with the latter insofar as our two selves are not sufficiently similar. Another way to put this point: the similarity constitutive of the third sense of identification is scalar, whereas numerical identity is not.

I believe this bare-boned account is quite promising. I also recognize that the account is still pretty bare-boned. “Sufficient” and “similarity” are all vague terms, for one thing, and I do not yet know how to precisify them in a satisfactory way. For another thing, my neutrality on what psychic elements are privileged in the volitional web (that is, the nature of the real self) prevents us from giving actual attributability conditions in any specific case, so the theory has of yet no practical payoff. Nevertheless, these are not so much faults as they are (gaping) gaps, challenges for those who find the formal aspects of the theory, such as it is, attractive.

There is one last point that may increase the attractiveness of the approach and it has to do with its causal component, which is vague, but deliberately so. Now it’s of course important to include some such causal requirement, because without it, one could find that certain actions are attributable to one solely in virtue of one’s utterly coincidental similarity to some stranger who performed them. The causal component at least rules out this possibility. One might think, though, that this component could then be met only in virtue of the relevant psychological states being manifested from their persistence in the same brain (or relevant part of the brain), but this would be too narrow a conception. I think it is important to allow for the possibility of my identification with other agents (even strangers), those who don’t share my brain but who nevertheless have had a significant causal impact on the formation of my real self, through my interaction with them or their ideas.[34] To the extent that their volitional selves have become mine by this causal route, then, so too have their actions. Of course, given our simultaneous existence, one could only say in such cases that attributability for their actions has been extended, not transferred, to me.

The implications of this point may now be clear. Not only may I identify with another agent, I may also identify with collections of agents, as long as they have a sufficient number of psychological elements in their volitional webs in common and the relevant similar features of my own volitional web are dependent on my causal interactions with them. When these other agents or their representatives engage in certain behavior, then, their actions may be attributable (as well) to me, rendering me eligible for accountability for their actions. This is the notion of distributive collective responsibility, or shared responsibility, a notion to which some theorists are attracted but which is itself very difficult to explicate.[35] The analysis given here may thus shed some light in this arena.

Regardless, though, of whether or not my schematic theory of identification and attributability is successful, we should not lose sight of the main point here: to understand the nature of Platitude, we must understand the nature of attributability, and while we may not yet be entirely sure of what it does consist in, we at least know what it does not consist in, namely, personal identity. Clearing Slogan from our path may thus afford us a fresh and unobstructed look at the actual conditions of this central feature of our moral practices.[36]

NOTES

REFERENCES

Bradford, J. and S.M. Smith. 1979. Amnesia and Homicide: The Padola Case and a Study of Thirty Cases. Bulletin of the American Academy of Psychiatry and the Law 7: pp. 219–231.

Bratman, M. 1992. Shared Cooperative Activity. The Philosophical Review 101: pp. 327–341.

—–. 1996. Identification, Decision, and Treating as a Reason. Philosophical Topics 24: pp. 1–18.

—–. 2002. Hierarchy, Circularity, and Double Reduction. In Contours of Agency: Essays on Themes from Harry Frankfurt, ed. S. Buss and L. Overton (Cambridge, MA: The MIT Press), pp. 65–85.

Brink,D. 1997. Self-Love and Altruism. Social Philosophy & Policy 14: pp. 122–157.

Buss, S. and L. Overton, eds. 2002. Contours of Agency: Essays on Themes from Harry Frankfurt. Cambridge, MA: The MIT Press.

Butler, J. 1975. Of Personal Identity. In Personal Identity, ed. J. Perry (Berkeley, CA: University of California Press), pp. 104.

Cima, M., et al. 2002. I Can’t Remember Your Honor: Offenders Who Claim Amnesia. German Journal of Psychiatry 5: pp. 24–34.

—–. 2004. Claims of Crime-Related Amnesia in Forensic Patients. International Journal of Law and Psychiatry 27: pp. 215–221.

D’Arms, J. and D. Jacobson. 2000. The Moralistic Fallacy: On the ‘Appropriateness’ of Emotions. Philosophy & Phenomenological Research LXI: pp. 65–90.

Darwall, S. 2006. The Second-Person Standpoint. Cambridge, MA: Harvard University Press.

Dancy, J., ed. 1997. Reading Parfit. Oxford: Blackwell.

DeGrazia, D. 2005. Human Identity and Bioethics. Cambridge: Cambridge University Press

Frankfurt, H. 1971. Freedom of the Will and the Concept of a Person. Journal of Philosophy LXVIII: pp. 5–20. Reprinted in Frankfurt 1988, pp. 11–25.

—–. 1988. The Importance of What We Care About. Cambridge: Cambridge University Press.

—–. 1999. Necessity, Volition, and Love. Cambridge: Cambridge University Press.

Glannon, W. 1998. Moral Responsibility and Personal Identity. American Philosophical Quarterly 35: pp. 231–249.

Haksar, V. 1980. Equality, Liberty, and Perfectionism. Oxford: Oxford University Press.

Helm, B.W. 2008. Plural Agents. Nous 42: 17–49.

Lewis, David. 1976. “Survival and Identity.” In The Identities of Persons, ed. A.O. Rorty (Berkeley, CA: University of California Press).

Locke, J. 1975. Of Identity and Diversity. In Personal Identity, ed. J. Perry (Berkeley, CA: University of California Press), pp. 39–51.

Madell, G. 1981. The Identity of the Self. Edinburgh: Edinburgh University Press.

Martin, R. and J. Barresi, eds. 2003. Personal Identity. Oxford: Blackwell.

May, L. and T. Raimo. 2007. Introduction. Journal of Social Philosophy 38: pp. 365–368.

Noonan, H. 1989. Personal Identity. New York: Routledge.

Olson, E.T. 1997. The Human Animal: Personal Identity Without Psychology. Oxford: Oxford University Press.

Parfit, D. 1984. Reasons and Persons. Oxford: Oxford University Press.

—–. 1986. Comments. Ethics 96: pp. 832–872.

Perry, J., ed. 1975. Personal Identity. Berkeley, CA: University of California Press.

Reid, T. 1975. Of Mr. Locke’s Account of Our Personal Identity. In Personal Identity, ed. J. Perry (Berkeley, CA: University of California Press) pp. 116-117.

Rorty, A.O., ed. 1976. The Identities of Persons. Berkeley, CA: University of California Press.

Scanlon, T.M. 1998. What We Owe to Each Other. Cambridge, MA: Harvard University Press.

Schechtman, M. 1996. The Constitution of Selves. Ithaca, NY: Cornell University Press.

—–. 2005. Personal Identity and the Past. Philosophy, Psychiatry, & Psychology 12: pp. 9–22.

Schoeman, F., ed. 1987. Responsibility, Character, and the Emotions. Cambridge: Cambridge University Press.

Sher, G. 2006. In Praise of Blame. Oxford: Oxford University Press.

Shoemaker, D. 2003. Caring, Identification, and Agency. Ethics 114: pp. 88–118.

—-. 2009a. Personal Identity and Ethics: A Brief Introduction. Peterborough, Ontario: Broadview Press.

—-. 2009b. Responsibility and Disability. Metaphilosophy 40: pp. 438–461.

—–. Forthcoming. Attributability, Answerability, and Accountability: Toward a Wider Theory of Moral Responsibility. Ethics.

Shoemaker, S. 1970. Persons and Their Pasts. American Philosophical Quarterly 7: pp. 269–285.

Smith, A.M. 2005. Responsibility for Attitudes: Activity and Passivity in Mental Life. Ethics 115: pp. 236–271.

Sider, T. 2001. Four-Dimensionalism. Oxford: Oxford University Press.

Strawson, P.F. 1963. Freedom and Resentment. Proceedings of the British Academy 48: pp. 1–25.

Striblen, C. 2007. Guilt, Shame, and Shared Responsibility. Journal of Social Philosophy 38: pp. 469–485.

Stump, E. 1988. Sanctification, Hardening of the Heart, and Frankfurt’s Concept of Free Will. The Journal of Philosophy 85: pp. 395–420.

—–. 1996. Persons: Identification and Freedom. Philosophical Topics 24: pp. 183–214.

Tadros, V. 2005. Criminal Responsibility. Oxford: Oxford University Press.

Thalberg, I. 1978. Hierarchical Analyses of Unfree Action. Canadian Journal of Philosophy 8: pp. 211–226.

Velleman, J.D. 2002. Identification and Identity.In Contours of Agency: Essays on Themes from Harry Frankfurt, ed. S. Buss and L. Overton (Cambridge, MA: The MIT Press), pp. 91–123.

Watson, G. 1975. Free Agency. Journal of Philosophy 72: pp. 205–220. Reprinted in Watson 2004, pp. 13–32.

—–. 2004. Agency and Answerability. Oxford: Oxford University Press.

Wiggins, D. 1967. Identity and Spatio-Temporal Continuity. Oxford: Basil Blackwell.

Wolf, S. 1987. Sanity and the Metaphysics of Moral Responsibility. In Responsibility, Character, and the Emotions, ed. F. Schoeman (Cambridge: Cambridge University Press), pp. 46–62.

—–. 1990. Freedom Within Reason. Oxford: Oxford University Press.

1. “Narrativity” 

If one is Narrative, then—here’s my first definition— one experiences or conceives of one’s life, one’s existence, oneself, in a narrative way, and in some manner lives in and through this conception. To be Narrative, as I will use the term, is to have a certain psychological characteristic.

It is in the first instance a natural disposition, even if it is open to cultivation. Narrativity, or the lack of it, is a natural dimension of human psychological difference, whatever the possible effects of training or cultural influence.

But what is it, exactly? What is it to experience oneself or one’s life, or major parts of one’s life, in a narrative way? This is a large question and I know of no clear answer. But there is widespread agreement that Narrativity exists. According to the neurologist Oliver Sacks, “each of us constructs and lives a ‘narrative’ … this narrative is us, our identities.”

The psychologist Jerry Bruner holds that “self is a perpetually rewritten story,” and “in the end, we become the autobiographical narratives by which we ‘tell about’ our lives.”

The philosopher Marya Schechtman claims that one “creates [one’s] identity by forming an autobiographical narrative—a story of [one’s] life.”

Charles Taylor holds that “wemust inescapably understand our lives in narrative form.”

“We are all storytellers,” according to Ruthellen Josselson, Amia Lieblich, and Dan McAdams, and “we are the stories we tell.”

David Velleman writes that “we invent ourselves,” in a paper called “The Self as Narrator,” adding that “we really are the characters we invent.”

According to Dan Dennett, we are all virtuoso novelists, who find ourselves engaged in all sorts of behaviour, and we always try to put the best ‘faces’ on it we can. We try to make all of our material cohere into a single good story. And that story is our autobiography. The chief fictional character at the centre of that autobiography is one’s self.

I think these quotations give the general idea. Most of them add the claim that one conceives of oneself—or one’s self—in a narrative way to the claim that one experiences one’s life in a narrative way, and make no clear distinction between the two things. I have accordingly built this into the starting definition of Narrativity.

All these writers endorse what I call

(a) the psychological Narrativity thesis

which states that all normal people are Narrative. All normal human beings conceive their lives, themselves, their existence, in a narrative way, and in some manner live in and through this conception. The psychological Narrativity thesis is an empirical, factual, descriptive thesis. This is how we are, it says, this is our nature.

Many also endorse a further thesis which embraces and extends the psychological Narrativity thesis. This is

(b) the narrative self-constitution thesis

which states that all normal people constitute their identity as persons or selves by virtue of being Narrative—by conceiving their lives, their existence, in a narrative way, and living in and through this conception. One of the most prolific exponents of this thesis is Dan McAdams, who propounds what he calls the “life-story theory of identity.”

“Identity,” he says, “is itself a life story.”

Schechtman puts it as follows: “we constitute ourselves as persons by forming a narrative self-conception according to which we experience and organize our lives.”

2. Are we all Narrative?

We have a major consensus. And it is arguable that there’s a way of understanding the psychological Narrativity thesis, at least, in which it is trivially true. And if it is trivially true then of course I endorse it. Truth is always good, trivial or not. But I am going to assume that it is meant to be non-trivially true, and consider the suggestion that it is false in any sense in which it is not trivial. As I understand the psychological term “Narrative,” some of us aren’t Narrative at all. Some of us do not naturally cast or construe our lives as a narrative or story of some sort, and we do not experience parts of our lives in this way either. Nor do we think of ourselves as opposed to our lives in a narrative way—whatever exactly this might be supposed to be. We may simply not be reflective about these things, or we may be like Bob Dylan:

I don’t think I’m tangible to myself. I mean, I think one thing today and I think another thing tomorrow. I change during the course of a day. I wake and I’m one person, and when I go to sleep I know for certain I’m somebody else. I don’t know who I am most of the time. It does not even matter to me.

or Samuel Hanagid a thousand years ago when he writes that:

for my part, there is no difference at all between my own days which have gone by and the distant days of Noah about which I have heard. I have nothing in the world but the hour in which I am: it pauses for a moment, and then, like a cloud, moves on.

The Narrativity theorists admonish Dylan: “Look, everyone is Narrative. Your ‘‘I am one thing today, another thing tomorrow’ thing is just part of your narrative about yourself.” And Bob Dylan has certainly told a lot of stories, when questioned about himself. But this, if anything, confirms the accuracy of his remark. As for the great Samuel Hanagid—spice merchant, warrior, poet, tax collector, grand vizier to the King of Granada—I think the Narrativity theorists should treat him with great care. At the very least, people vary greatly when it comes to the basic Narrative impulse. If you consider your friends for a while, I think you will see how they differ in this respect.

The Narrativity theorists may concede that some people do not appear to be Narrative, but claim that they are really; it is just that they are doing it unconsciously or implicitly. The Narrativity theorists may back up their claim by instancing the deep human drive to make sense of things, find patterns and explanations at all costs, and, where necessary, “confabulate” in the technical sense of experimental psychology: that is, make up, invent, patterns and explanations where there are none to be found.

Here again, though, we have to do with a central human “individual difference,” a psychological variable, a trait that is strong in some and weak in others. Keats congratulated Shakespeare on his “Negative Capability,” his capacity to rest in “uncertainties, Mysteries, doubts, without any irritable reaching after fact & reason” (the sort of irritability that can easily lead to confabulation).

I think some of us possess a good  amount of negative capability when it comes to our attitude to ourselves and our own lives. Others could probably do with more of it. It may be sounder—closer to life—than the perilous reachings of narrative self-interpretation, let alone narrative self-constitution. We are perhaps, in certain ways, and to a deep degree, “strangers to ourselves”, in the words of the social psychologist Timothy Wilson.

In many respects, it seems, we do not know what we are about. Perhaps we begin to know something about ourselves when we begin to appreciate this. Whether this is so or not, any large-scale project of self-constitution, narrative or not, is likely to involve a fantasy of control. It risks being violently Procrustean. “If the hare has seven skins,” as Nietzsche says, then “the human being can shed seven times seventy skins and still not be able to say: ‘This is really you, this is no longer outer shell’.”

Nietzsche may be exaggerating, and I hear a voice—Marya Schechtman’s—telling me that there are things that people mean by the expression “narrative self-constitution” that I cannot possibly wish to reject. That may be so. For the moment, if the Narrativity theorists insist that we are all really Narrative, I will disagree vehemently, and again begin to wonder whether they are defining Narrativity in such a way that the psychological Narrativity thesis comes out as trivially true. When McAdams, Josselson, and Lieblich say that ‘we are all storytellers … what is consciousness but an inner narration of experience?’, they seem to understand narration in a way that immediately lands them in triviality. It seems that to be conscious at all is in Kant’s phrase ‘always already’ to be engaged in narrative.

So we are all Narrative, simply because we are all conscious (except for the zombies among us).

On one view, having and making plans already entails having a Narrative outlook and engaging in narrative self-constitution. In this case triviality threatens again, for we all make some plans. Some have no long-term plans because they are feckless or shiftless, or, more positively, because they are highly spontaneous or happy-go-lucky (“very improvident and cheerful,” like the Flopsy Bunnies).

Others make no long term plans because they are desperately poor and are simply trying to survive from day to day. But we all make some plans. So we are back with triviality—and a continuing unclarity about what exactly the psychological Narrativity thesis and narrative self-constitution thesis amount to.

The unclarity that derives from the key term in the definition of Narrativity: “narrative”. What is to be done? Aristotle steps in with his usual reminder: do not attempt more precision than the subject matter admits. But the need for clarification remains. It does not help that most uses of the term “narrative” in the humanities can be replaced without semantic loss by words like “description,” “account,” “view,” “theory,” “explanation,” ‘understanding,” “belief,” “concept,” “picture.”

Perhaps 95 per cent of all uses of “narrative” are replaceable in this way, once we put literary theory aside, inside or outside the humanities.

Any description of anything whose existence involves change is liable to be called a “narrative.” And perhaps change is not necessary—as in the case of the description of a painting.

This suggests a rule of engagement. Whenever anyone uses the word “narrative,” ask whether it can be replaced without semantic loss by by one of the words listed above, or by some variant of the generic expression “description or account of a temporally extended sequence of events.” If it can be replaced, let it be replaced. Then we will be able to see more clearly what is left. If it is hard to say whether or not something is lost in the replacement, the burden of proof (proof that something has been lost) will fall on the user of the word narrative.” If we proceed in this way, we may be able to get more clarity into the debate. It will also help if we explicitly restrict attention to the use of the word “narrative” in the ethical and psychological contexts that principally concern philosophers. We will still need to make a direct assault on the question of the intended force of the word, but we will have a much firmer base from which to start.

This, though,is a task for another time. Having marked the need for clarification, I am going to rely on the characterization of narrative contained in the quotations I’ve already given, in opposing the psychological Narrativity thesis and the narrative self-constitution thesis, along with their ethical counterparts, which I have not yet introduced. 

3. Difference

The key claim is simply that human beings differ. We differ dramatically in respect of Narrativity. There are the Dan Dennetts, on the one hand, spinning away, trying for whatever reason to make all of their life material “cohere into a single good story,” and the Bob Dylans, on the other, who are doing no such thing. There are people like Charles Taylor who hold that a “basic condition of making sense of ourselves is that we grasp our lives in a narrative,” and have an understanding of our lives “as an unfolding story” (1989: 47, 52), and people like myself who find the idea of engaging in this sort of self-narrativizing activity profoundly alien, people who believe that they can get on quite well without it, whether or not they also believe that it risks being a royal road to error. We differ in respect of Narrativity as much as Gerard Manley Hopkins and Iris Murdoch differ when it comes to their sense of themselves in the present moment. Hopkins speaks of his “self-being”:

my consciousness and feeling of myself, that taste of myself, of I and me above and in all things, which is more distinctive than the taste of ale or alum [or] the smell of walnutleaf or camphor, and is incommunicable by any means to another man.

Iris Murdoch’s husband John Bayley reports that

Iris once told me that the question of identity had always puzzled her. She thought she herself hardly possessed such a thing, whatever it was. I said that she must know what it was like to be oneself, even to revel in the consciousness of oneself, as a secret and separate person …. She smiled, looked amused, uncomprehending.

Proponents of “virtue ethics” tend to say that there is one right way to be. Aristotle claims that we should before all else possess the four “cardinal” virtues of courage, temperance, “practical wisdom” and justice. And these are surely all very good things to have. Even here, though, we need to recognize the fact and importance of human difference, the very different forms that human virtue can take given the vast field of human imperfection. There seem to be people whose virtues are forms of their failings. There are people whose gifts depend on their faults. There are people who inspire love and respect although they are rackety, partial, inconsistent and comically faint-hearted (as opposed to temperate, just, practically wise, and courageous). There are on the one hand heroes and wise maidens with oil lamps, and, on the other hand, Don Quixotes, Fitzcarraldos, Shirley Maclaine figures, Mullah Nasruddins, Uncle Tobys, and Papagenos. There are people whose charm is inextricable from their rashness or naughtiness, people whose creative energies are constitutively linked to their cowardice or bad temper, people whose particular insight and humour depend on certain sorts of injustice and unfairness, people whose special capacities for generosity and great-heartedness are inseparable from their intemperance, hopelessly faulty people who have moral charm (a notion favoured by Isaiah Berlin)—moral charm which constitutively is bound up with their lack of practical wisdom. One should not be too quick or grand when one tried to generalize about the human condition—either about how it is or how it ought to be.

4. Two more theses

Sacks, Bruner, McAdams and many others endorse (a) the psychological Narrativity thesis and (b) the narrative self-constitution thesis. We are all Narrative, and we “constitute ourselves as persons by forming a narrative self-conception according to which we experience and organize our lives.” Beginning in late adolescence and young adulthood,

we construct integrative narratives of the self that selectively recall the past and wishfully anticipate the future to provide our lives with some semblance of unity, purpose, and identity. Personal identity is the internalized and evolving life story that each of us is working on as we move through our adult lives …. I … do not really know who I am until I have a good understanding of my narrative identity.

I find this bewildering. I certainly have an identity in the relevant sense, a psychological identity, and I certainly have a past, a history; even, if you like, a life story. But I am blowed if I constituted my identity, or if my identity is my life story.I do not spend time constructing integrative narratives of myself or my self that selectively recall the past and wishfully anticipate the future to provide my life with some semblance of unity, purpose, and identity. When I read that this is a human universal, I think I must come from the planet Zog.

Some psychologists use the term “identity” to refer not to how one fundamentally is, psychologically, but to how one conceives oneself to be—however misguided one is. Some also use the term “self” in this way. On this view, one’s self is just one’s self-conception. One’s identity, one’s actual psychological identity, is how one conceives of oneself. This seems to me like saying that one’s VW is a Rolls Royce (or conversely), because that is what one thinks it is. It also sounds rather alarming. But this, perhaps, is just a terminological matter; and since it seems that McAdams is using the term “identity” in this special way, the disagreement between the two of us is considerably less than it seems. I cannot, for example, object that his view seems to have the consequence that no one can ever be self-deceived.

Even so, disagreements remain—among them, the deep disagreement between those who think that one’s self-conception is essentially some sort of narrative and those who do not. I am also unsure what to make of McAdams’s statement that ”I … do not really know who I am until I have a good understanding of my narrative identity.” The trouble is that the phrase “know who I am” seems to me to connect inexorably to an “objective” understanding of the word “identity” to mean who I really am as a matter of deep psychological fact, opposed to the “subjective” use of “identity” to mean how I conceive myself to be. It seems all too painfully clear that I may have a very good understanding of my identity in the sense of my account of myself to myself, and not “know who I am” at all. I may be clueless about myself, radically self-deceived as to my identity in the objective sense, and my cluelessness may consist precisely in my identity in the subjective sense, my McAdamian identity.

But having noted this difficulty, and possible ground of misunderstanding, I am going to put both aside for now. I will continue to use “identity” in the way I favour, as referring to who or what I am, fundamentally, considered as a person, however much or little I know about the matter.

A person can of course be highly complex and inconsistent compatibly with their being a fact of the matter about how they are, fundamentally, considered as a person.

(a) and (b), the psychological Narrativity thesis and the narrative self-constitution thesis, are distinct. One can endorse (a), the view that we are all Narrative if normal, without in any way endorsing (b), the view that we (all normally) constitute our identity in that way. (a) is also independent from any other self-constitution view—any view according to which we constitute ourselves as persons in some way. Neither thesis has any evaluative implications; both are merely empirical and descriptive. But they both have evaluative or “normative” forms. I call the normative version of the psychological Narrativity thesis

(c) the ethical Narrativity thesis.

According to the ethical Narrativity thesis, one ought to have a Narrative outlook on one’s life. One ought to think about oneself and one’s existence in a narrative way and in some manner live in and through this conception. It is essential to a good life, essential to living well, essential, in fact, to true or full personhood.

We can illustrate this view by supposing that Socrates is right when he says that the unexamined life is not a full human life, not a life for a human being—right that we ought to be reflective about our lives in some way or other. I am not sure that this is a universal human truth, in fact, but let us assume that it is (“in some way or other” may prove to be a fairly capacious dispensation). We may then take the ethical Narrativity thesis to be the thesis that we ought to reflect about our lives in a particular way: in the narrative way, in the narrative framework.

This view is widely held today, and it is usually part of the normative version of the narrative self-constitution thesis, which I will call (somewhat heavily)

(d) the ethical narrative self-constitution thesis

According to this view, one’s identity as a person or self ought to be constituted by (or perhaps in or through) one’s narrative of one’s life.

Like (a) and (b), (c) and (d), the ethical Narrativity thesis and the more specific ethical narrative self-constitution thesis, are distinct. For one can endorse (c), the claim that one ought to be Narrative, without endorsing (d), the more specific claim that one ought to constitute one’s identity in a narrative way. In practice, though, most of those who think that one ought to be Narrative seem to think that this is at least partly, if not wholly, because one ought to constitute one’s identity in a narrative way.

We now have four connected theses: a thesis about how we do normally think about ourselves and our lives; a thesis about how we ought to do this (given that Socrates is right that we ought to do it in some way or other); a thesis about how our identity gets constituted; and a thesis about how it ought to get constituted.

It may be said that the fundamental features of one’s identity are what they are independently of anything one can do about them. There is, first and foremost, the matter of a person’s position in the great state-space of core character traits, which once had four main parameters—sanguine, melancholy, phlegmatic, and choleric—and now has five, the “Big Five,” the bipolar categories Extraversion/Intraversion, Neuroticism/non-Neuroticism, Agreeableness/Disagreeableness, Conscientiousness/Irresponsibility, Openness/Closedness to Experience. Well, this is surely right, and I will return to the point. For the moment it is enough to note that one can ingest and savour this large dose of realism—highly restorative in a debate which sometimes runs wild—while continuing to endorse versions of the two self-constitution theses which remain highly substantive, although much more moderate than their unrestricted versions, in focusing on self-constitution in so far as as there is anything we can do about ourselves and our identities.

5. Four initial positions

I want now to put aside the specific issue of self-constitution (theses (b) and (d)) for a while, and consider the four possible theoretical positions defined by acceptance or rejection of (a) and (c), the psychological Narrativity thesis and the ethical Narrativity thesis.

First position. The descriptive psychological thesis is true but the ethical thesis is false. That is, we are indeed deeply Narrative in our thinking about ourselves, but it is not a good thing. It leads to self-deception, bad faith, inauthenticity, bullshit. This view is associated particularly with Jean-Paul Sartre, but we may also call the great (second-century) Stoic Marcus Aurelius as a witness.

Second position. The descriptive thesis is false, but the ethical thesis is true. It is not true that all normal people are naturally Narrative in their thinking about ourselves. But we should all be Narrative. We need to be, in order to live a good life. Bob Dylan needs to start narrativizing before it is too late. This is a common view today, and it also has respectable ancestry, e.g. in the person of (the first-century historian and philosopher) Plutarch.

Third position. Both theses are true: both the psychological Narrativity thesis, according to which all normal non-pathological human beings are naturally Narrative, and the ethical Narrativity thesis, according to which being Narrative is crucial to a good life.  This seems to be the most popular view in the academy today, and indeed in the clinic, followed by the second view. It does not entail that everything is as it should be; it leaves plenty of room for the idea that many of us would profit from being more Narrative than we are, and also, I take it, for the view that we can narrativize ourselves badly, and learn to do it better.

Fourth position. The psychological Narrativity thesis and the ethical Narrativity thesis are both false. This is my view. I do not think it is true that there’s only one general way in which (non-pathological) human beings think about themselves when they think about themselves in the larger, more overarching way that is at present in question—the Narrative way.  And I do not think there’s only one good way for them to do so— the Narrative way. There are deeply non-Narrative people and there are good ways to live that are deeply non-Narrative. There are people who live well who are not only not naturally Narrative in temporal temperament, where this is something merely negative; there are people who are naturally strongly anti-Narrative, where this is something positive in its own right. People like this not only lack any inclination to cast, conceive, construe and conduct their lives (or parts of their lives) in a Narrative way; they also find it peculiarly hard to do. They are likely to stumble and stall in recounting events in their lives, and this is principally because they have a strong sense that something is being positively misrepresented, falsified, in so far as the events are being strung together in a narrative.

6. Two more positions

With these four basic positions in place, we can return to the narrative self-constitution thesis, along with its normative version. The non-normative version, recall, says that your narrative of your own life constitutes your identity as a person or self. The normative version says that one’s identity as a person or self ought to be constituted by one’s narrative. The dominant view in the academy and the clinic, I think, is that both these theses are true. Proponents of this view hold a full house: we are naturally Narrative; we ought to be; we constitute our identity in this way; and we ought to. I am at the opposite end. We are not all naturally Narrative; it is rarely if ever a good thing; we do not all constitute our identity in this way; nor ought we to. I suspect that Michael Jackson was destroyed by his belief in narrative self-constitution.

Who iss right, and why does it matter? It matters both theoretically and practically. The psychological Narrativity thesis (according to which we are all Narrative) is an empirical claim whose truth or falsity is of great moment to anyone interested in human affairs. The ethical Narrativity thesis (we all ought to be Narrative) is a normative, evaluative claim, not a descriptive claim, but it also of course makes a claim to truth, an empirical claim, in fact, about what is actually good for people, and one reason why it matters is that if it is not true then a considerable amount of damage may be being done—in psychotherapeutic contexts, for example—by people who think it is true. Therapy aside, and more generally, the ethical Narrativity thesis creates a problem inasmuch as non-Narrative people who are exposed to it may think there’s something wrong with them and their lives when there is not.

One obvious possibility is that the Narrative life is right or best for some but not for others. It may be, for example, that naturally Narrative people ought to be Narrative—that this is a good way for them to lead their lives—while naturally non-Narrative people ought not. I am certain that the Narrative life is not good for everyone, and I am interested in the polemical thesis that it is not a particularly good way—or at least the best way—for anyone to live their lives. Perhaps it is essentially confining, essentially petty, excessively self-concerned, second-order, shopkeeperly. Perhaps the ethical Narrativity thesis is false for everyone, or almost everyone, because we live beyond any tale that we happen to enact, and the ethical Narrativity thesis seeks to confine us to our tales. When I first wrote about Narrativity I wanted to defend the non-Narrative or anti-Narrative temperament against what I took to be the prevailing orthodoxy, and I thought that the most important thing to do was to stress human difference and variety, as—so far— I have done again in this paper. Now I want to begin to go on the offensive against Narrativity. If the Narrativity thesis is the main thesis in town, we need the antithesis, the anti-thesis. (Aufhebung can wait.) I am going to proceed in two stages.

7. Hexetasis 

Stage 1. I have allowed for the sake of argument that Socrates is right that we ought to think about or examine ourselves in a wide, general, life-assessing way. I have allowed that he is right that we ought to engage in what he called hexetasis, or more accurately autohexetasis—self-examination (I am going to use the Greek word, because it is unfamiliar and to that extent theoretically unencumbered). Question: Does autohexetasis, hexetasis for short, essentially involve narrative, Narrativity? If I have to be hexetatic, do I have to be Narrative? Is the Narrative way the best way, even if it is not the only way? I say No and No. Let me put down a clear marker straight away by saying that I do not think hexetasis involves Narrativity even when it involves therapy that pays special attention to past life, especially early life.

Perhaps this shows that I understand the notion of narrative is a narrower way than others. So be it. I do not think that deriving therapeutic benefit from grasping and emotionally appreciating causal connections between one’s early-life experiences and one’s current travails need involve any sort of specifically or distinctively Narrative thinking—any more than learning that the scar on one’s finger is due to an early fall. In most cases, I think therapy delivers a strikingly disparate clutch of insights, bits and pieces, elements of understanding (emotional understanding) that cast light severally on this and that. The insight-elements do not themselves make up a narrative sequence; nor is the way each one casts light individually a distinctively narrative matter. Truth in this area is mostly a matter of fragments and oddments, accurate hexetasis is largely bricolage, smooth narrative is almost certainly fantasy (except, perhaps, for people who are obsessional in one way or another). I certainly do not want to rule out the possibility that successful therapy or hexetasis should in some cases have a narrative form. But if we measure therapeutic success partly by reference to getting at the truth, rather than just by reference to restoring or increasing everyday fitness for life, then I think the possibility of achieving success with smooth narrative grows slimmer.

No, say the narrativists, the facts are these. Almost all causal explanation of features of concrete reality—in this case, features of oneself, one’s personality, one’s present situation, and so on—involves identifying links between past and present. And when one’s subject is oneself and one’s life, all identification of links between past and present involves narrative. Granted, they say, this may not be true (or interestingly true) in the case of many explanations of features of oneself, medical and genetic explanations, for example, to do with eyesight and shoe size, but it is going to be true in the case of many of  the most interesting character-related or more generally psychological explanations. This means that hexetasis is bound to involve narrative as soon as it involves causal explanation, past-present linking. And it is bound to involve a considerable amount of highly substantive causal explanation if it is to get anywhere. So it must after all involve a considerable amount of highly substantive narrative.

In sum and in brief, we have an argument. Premise [1]: Hexetasis requires causal explanation, past-present linking. Premise [2]: Causal explanation is necessarily narrative, when one’s subject is oneself and one’s life, and in particular one’s ethical life, in the largest sense of that large word. Conclusion: hexetasis is necessarily narrative.

The argument is valid, but I do not think it is sound. I reject both premisees. As for premise [1], I do not think hexetasis necessarily involves causal explanation; I think, with Proust, among others, that it often involves such simple things as discovering something about what really matters most to one (I will come back to this). But although I do not think that hexetasis necessarily involves causal explanation, I do think that it very often involves causal explanation. So I am prepared to grant premise [1] for the sake of argument. But not premise [2]. I reject premise [2]: causal explanation, psychologically significant causal explanation, is not always narrative. It is not ”always already” narrative, even when a life is in question. It can be, in an intuitive sense, anti-Narrative. I can have no story of development, but capture an isolated key explanatory connection between past and present, an arc of contact that reaches above and is utterly independent of all the developmental, progression-involving details of my life.

8 “Le vrai moi”

So much (for the moment) for hexetasis. Now for a frontal assault on the “full house” view described on p. 000. I have a team of supporters. I cannot write like Montaigne, but I can match him in appealing to others and moving from quotation to quotation. On my team I have, for a start, or so I believe, people like Proust, Emerson, Woolf, and at least one avatar of Nietzsche, who has, it seems as many avatars as the Hindu God Vishnu (at least one of which is on the other team). This initial list may sink some hearts, on the grounds that these people are just too sensitive, too highly strung, too transcendental-intellectual, too hyper. But they are counterbalanced by the comparative bluntness of some of the other team members: the great short-story writer V. S. Pritchett, the source of my title, and the incomparable Chekhov, who inspired Pritchett’s observation.

Kierkegaard has also been claimed by both sides in the Narrativity debate, but he starts well for me by saying that one should not try “to arrange oneself dramatically in [time or in] temporality,” for this sounds like a direct criticism of Narrativity, an exact rejection of the MacIntyre/Taylor project, and also, I think, the Ricoeur project, although Ricoeur’s formulations seem somehow gentler and more open than MacIntyre’s or Taylor’s. Kierkegaard’s most famous remark on this subject runs as follows:

Philosophy is perfectly right in saying that life must be understood backwards. But then one forgets the other clause—that it must be lived forwards. The more one thinks through this clause, the more one concludes that life in temporality never becomes properly understandable, simply because never at any time does one get perfect repose to take a stance: backwards.

This passage has probably been cited in support of the Narrative cause. It seems to me, though, that it rejects the idea that one should make use of a story of oneself in living forwards, that is, in living—for there is no other way to live. Your narrative is something you work out, perhaps, if you want to, looking back, but you should not let it dictate your stance to the future. Rather the contrary, when it comes to dictation. In another passage Kierkegaard says that we should live like someone taking dictation, continually having his pen poised for what comes next, so that he does not presume meaninglessly to place a period before the meaning is complete or rebelliously throw away his pen.

This is, admittedly, in context, a counsel against despair, but it seems, more generally, a strongly anti-Narrativist injunction.  I do not, however, want to rest anything on Kierkegaard’s case, which is highly complex, particularly when it comes to the importance he accords to reflective recollection (as opposed to mere memory).

Proust considers days when one finds oneself “outside the regular tenor of life” (“en dehors du train courant de la vie”), days when one is more inward. They are not days on which one finds oneself in a special position to “appreciate what one has made of oneself.” They are days when one is simply more at home to what one just is, independently of any making, more at home to—or with, or in—“le vrai moi,” as Proust calls it, the true me or true self, to which there is, on these days, some kind of reconnection, or better connection. One has been caught up in one’s daily life (or one’s narrative?) in a way that has in some respect alienated one from oneself. Perhaps one has been caught up in one’s narrative in a way that has alienated one from oneself. How could this be, the narrativists ask? Because we live beyond any tale that we happen to enact.

Proust gives great weight to what he calls “involuntary memories,” upsurgings of past episodes of extreme happiness. They teach you about yourself as nothing else can. They are rare and fundamental data. These data come from the past, to be sure. It does not follow that they have anything to do with narrative. On the contrary; they are isolated clues, clés, not clews you have to ravel up (narratively, temporally) to their source. They are complete in themselves in respect of their probative force. Proustian self-knowledge centrally involves things like this, suddenly realizing that one really loves something, something which is likely to seem wholly trivial from most points of view. Proust’s greatest revelations come from the feel of a starched napkin, the sensation of uneven flagstones under his feet, hearing a spoon clink on glass, dipping a biscuit in tea. Such things are, perhaps, the fundamental clues to self, under all one’s skins.

Of which there are at least 490, as Nietzsche says before he goes on to make the very point Proust is making. He, Nietzsche, is speaking of the dangers of self-examination, but he then says that there is a way in which this absolutely crucial enquiry [into oneself] can be carried out. Let the young soul look back upon its life and ask itself: what until now have you truly loved, what has raised up your soul, what dominated it and at the same time made it happy?

Again, this hexetatic use of the past is not a matter of narrative. It is a leap outside narrative. Obviously in looking back one considers parts of one’s actual life (one’s actual history), but what one hopes to acquire is information about one’s deep form, information about one’s nature that is not constituted—perhaps not even partly constituted—by the events of one’s actual life, let alone one’s narrative of that life, information that may seem oddly trivial and thoroughly surprising.

Emerson (deeply admired by Nietzsche) adds this, in his roaring essay on self-reliance:

What is the aboriginal Self…? The inquiry leads us to that source … which we call Spontaneity or Instinct … [or] Intuition. Every man discriminates between the voluntary acts of his mind, and his involuntary perceptions, and knows that to his involuntary perceptions a perfect faith is due. He may err in the expression of them, but he knows that these things are so, like day and night, not to be disputed. All my wilful actions and acquisitions are but roving; — the idlest reverie, the faintest native emotion, command my curiosity and respect.

Here Emerson delivers a colossal blow against many who champion self-constitution by wilful choice and action, but this is only en passant. His principal message is that we are reliably accessible to ourselves only by intuition … we need to listen for ourselves … we are in Nietzsche’s phrase “pieces of fate” … we wander most blindly precisely when we try to regiment ourselves … things are not up to us … we ourselves are not up to us.

He takes the point to its fiercest conclusion:

Nothing is at last sacred but the integrity of your own mind …. I remember an answer which when quite young I was prompted to make a valued adviser, who was wont to importune me with the dear old doctrines of the church. On my saying, ‘What have I to do with the sacredness of traditions, if I live wholly from within?’ my friend suggested — ‘But these impulses may be from below, not from above.’ I replied, ‘They do not seem to me to be such; but if I am the Devil’s child, I will live then from the Devil.’ No law can be sacred to me but that of my nature. Good and bad are but names very readily transferable to that or this; the only right is what is after my constitution; the only wrong what is against it.

If this seems too much, remember that Emerson is betting on the basic goodness (or good-enoughness, in Winnicott’s terms) of untainted human being. Kant, by contrast, believed in radical evil, that is, a tendency to wrongdoing right at the root [radix] of human being. Kant’s standards, though, were very high, and Emerson is at one with Kant in his paean to autonomy when he says that “no law can be sacred to me but that of my nature”— even as he utterly rejects what Kant has in mind, and, as I read him, reverses the aspirations of those in the present day who seek autonomy in self-constitution.

Both Emerson and Kant agree with Nietzsche that “you should become what you are.”

—“You cannot fail to become what you are. It is tautological that you are at any stage of your life what you have thus far become.’”

Emerson’s and Nietzsche’s point is precisely that it is all too possible that you can fail to become what you are, and miss out on reality, “that reality which we run a serious risk of dying without having known, and which is quite simply our life, true life.”

—“Of all the things you could have quoted in support of your argument, this is the worst. For Proust’s very next sentence reads as follows: ‘True life, life finally uncovered and illuminated, hence the only life that is fully lived, is literature.’”

What Proust means by this, though, is the polar opposite of what the narrativity theorists have in mind when they encourage us to be Narrative, to engage in narrative self-constitution. He means, very briefly, a certain high quality of apprehension  of life and self that is free from the almost invincible distortions and blindnesses of train-train habit, and fundamentally independent of the actual course of your life, and a fortiori of any narrative account of it.

Proust might nonetheless be thought to believe in self-constitution. For he also says that “we work away constantly trying to give our life its form.” But what happens when we do this, he says, is that “in spite of ourselves we copy, as in a drawing, the traits of the person we are, not the person we would like to be.”

This is not really self-constitution, and even if it were it would not be distinctively narrative self-constitution (and Proust seems here to have our public history and social persona principally in mind).

That is enough Emerson, Nietzsche and Proust. I am sure we should get in touch with ourselves (although I am not sure how this fits with projects of radical self-constitution). But doing so is often more a matter of luck than than endeavour, and I am now interested in a quieter point. For Mrs Ramsay, in Virginia Woolf’s To the Lighthouse, … it was a relief when they [the others] went to bed. For now she need not think about anybody. She could be herself, by herself. And that was what now she often felt the need of—to think; well not even to think. To be silent; to be alone. All the being and the doing, expansive, glittering, vocal, evaporated; and one shrunk, with a sense of solemnity, to being oneself, a wedge-shaped core of darkness, something invisible to others. Although she continued to knit, and sat upright, it was thus that she felt herself.

Here, I think, we have contact, contact with one’s identity, the distinguished thing in both senses of the word (though not in Henry James’s meaning). And it seems to be the negation, the pure negation, of narrative.

Consider, also, Woolf’s “moments of being.”

Some Woolfian moments of being are good but relatively ordinary moments of life in which one is—manages to be—present in a way in which one is often not present.

Others are memories, not necessarily happy (they may be ‘sledge-hammer’ blows), that have Proustian force. These are, as it were, windows into one’s essence, one’s personal essence—something to whose existence I am now happy to commit myself as a matter of basic psychological realism, without saying that we should go digging after it, or that it is unalterable, or that experience after early childhood can have no part in its constitution. Moments of being reveal something; something is made manifest.

—“The metaphor of the window is not neutral, given the current debate. It is potentially question’begging: it presupposes a fixed place—a kind of shrine—into which one looks.”

It needn’t be taken so crudely.

—“But these ‘moments of being’ are obviously part of one’s ‘narrative’, in the sense of one’s account of one’s life.”

This threatens to be question-begging in its turn. Or rather, it threatens to trivialize the claim that all normal people are Narrative. One cannot be said to have a narrative of one’s life, a narrative attitude to one’s life, a narrative self-conception, simply because one has and remembers and treasures such moments. We all have histories, and nearly all of us have a reasonable number of memories of past events in our lives. This alone cannot make it true to say that we are Narrative—short of trivializing the claim. What distinguishes Woolf’s moments of being, it seems to me, is their isolation from narrative. They are, certainly, actual events in her life, but that does not place them in a narrative. They are loci or manifestations of her fundamental form, or aspects of her fundamental form. Their power lies in their stillness, their self-sufficiency, their independence of the flow of life, their immunity to “making.” They make me think of the stillness of a great portrait. There is a fundamental mode of human understanding in which to understand someone is like experiencing this stillness.  The same goes, I propose, for understanding oneself. The correct image, in any case, and independently of the image of the portrait, is one of discovery, not creation or constitution.

Thoughts of this kind are no doubt familiar enough; the gathering conclusion is plain.  Theses (a)–(d) are all false. (a) and (b) are empirically false as descriptive theses. (c) and (d) are empirically false as normative theses. But suppose you are naturally Narrative, suppose you do naturally think of yourself and your life in a narrative way. Many do not do this, but suppose you do. So be it. That is one natural way of human being. But it seems to me that you should then be careful what else you do. For to identify oneself with one’s narrative of oneself, to take one’s narrative of oneself to constitute one’s identity, is likely to be a desperately reductive (if not murderous) act— and for several reasons. Not only because we live beyond any tale that we happen to enact, but also because there are many respects in which we are not very good at knowing much about ourselves, or about why we do what we do. To all the broadly speaking Freudian evidence in support of this last claim we have to add all the situationist evidence from social psychology. Then we have to take full account of the non-Freudian notion of the “adaptive unconscious” as expounded, for example, by Timothy Wilson in his book Strangers to Ourselves. The adaptive unconscious is the unconscious mind as acknowledged in cognitive science, a vast repertoire of mental processes that influence our judgments and decisions in ways inaccessible to conscious awareness. It turns out to do a great deal of the running of one’s life, especially when things go well, sometimes with little regard for one’s conscious elucubrations.

Note that this does not mean that it is not really one oneself who acts, when one acts. To think this is to make the great mistake, to identity oneself with the “conscious self,” which is, at best, the spindrift of your being. It just means that you’re not generally au fait with what is going on in the way that you think. And the danger is that the more you identify with your explicit narrative of yourself, the less you’re likely to know what is really going on. “We do not deal much in fact when we are contemplating ourselves”, as Mark Twain observed.

Some of us, do, perhaps, but most of us do not. “It is all very well to be aware of your awareness,” Lewis Thomas adds, “even proud of it, but do not try to operate it. You are not up to the job.”

The consequences for the (b) and (d), the ethical Narrativity thesis and the ethical narrative self-constitution thesis are I think clear. I am inclined to adapt William Blake’s famous quatrain: “Never seek to tell thy life, / Life that never told can be; / For the gentle wind doth move / Silently, invisibly.”

None of this is to say that one cannot plan a life’s work or pursue great projects. It is not to say that one cannot try to be good, to become good. (For that, though, one need never look further than the present; to think of yourself as on a “quest” is almost inevitably corrupting.) It simply raises doubts, grave doubts, in my opinion,  about the value of being Narrative and about the goal of narrative self-constitution, or indeed any project of self-constitution—about the extent to which self-constitution is likely to be a form of self-abuse—about the extent to which the goal of autonomy is a form of enslavement. As Pritchett says in his essay about Chekhov’s short stories, which surely inspired him, “life is a fish that cannot be netted by mood or doctrine, but continually glides away between sun and shadow.”

9. Conclusion  I’ve tried to present an antithesis to the Narrativist thesis. There’s a great deal more to say about all this, but this is as far as I’ve got at present.

1. The Explanatory Role of Experience

Berkeley (1734) thought that we cannot but take the existence of a thing to be a matter of its being perceived: “The table I write on, I say, exists, that is, I see and feel it.” In general, he thought that we could not form the conception of an objective, mind-independent reality: “all the choir of heaven and furniture of the earth, in a word all those bodies that compose the mighty frame of the world, have not any subsistence without a mind … their being is to be perceived or known” (Berkeley 1734). The argument that seems to compel Berkeley to this conclusion is, I think, quite forceful. The problem it raises is how we can acknowledge its force while not being swept along to his conclusion. The argument that is really doing the work in Berkeley is an argument from the explanatory role of experience. There are two basic ideas in the argument:

(1) Our understanding of concepts of the medium-sized world is grounded in our sensory experience, and
(2) Sensory experience can provide only concepts relating to sensory experience itself.

More generally, we might say that what is driving Berkeley’s argument is the idea that perceptual consciousness has an epistemic role to play in our cognitive lives. That is, it is because we are conscious that we have knowledge of particular properties and objects. It is only because we are conscious that we so much as have the conception of the various characteristics and objects that are in our world. And it is only because we are conscious that we have perceptual knowledge of the facts about our world. But, Berkeley is arguing, the only epistemic role that experience can have is to provide us with knowledge of experience itself. Berkeley’s argument is troubling and forceful for us because each of these premises has a lot of plausibility. Certainly, each of these premises would find many supporters among contemporary philosophers. Suppose you simply ask what the point is of perceptual experience, what good it does us that we have sensory experience. The answer that immediately seems compelling is that it is because we have sensory experience that we know about the things around us and their characteristics. It is only because of our experience of ordinary objects and their properties that we so much as have the conception of those objects and properties. At the same time, suppose you ask, “How exactly can sensory experience provide us with knowledge of the things around us?” People very often think of consciousness as a matter of experience having a certain qualitative character; and how can that qualitative character of experience be a crucial element in our knowledge of anything other than that very type of experience itself? The qualitative experiences, the qualia that we find in our minds, do not provide us with insight into what the external world is like. There are no qualia in any non-mental, physical world. It is hard to see how consciousness as such can provide us with knowledge of anything but itself. So if we accept that it is our experience of ordinary objects that allows us to form the conception of those ordinary objects, we seem driven to the conclusion that we can form the conception only of objects that are themselves merely aspects of our experiences, objects that are themselves creatures of the mind. Berkeley’s Puzzle is to understand how it could be that both (a) our grasp of concepts of ordinary objects is grounded in our experience of those objects, and (b) we have the conception of mind-independent objects. To resolve this Puzzle, I will argue, we have to look again at our conception of sensory experience; we have to think of sensory experience not as a matter of having qualia, for example, but as a matter of our being related in the right kind of way to the objects and properties around us.

There are two radical resolutions of the Puzzle. One is to say that the conclusion is correct: that we have only the conception of experience itself, and no such idea as that of a mind-independent material object. This is Berkeley’s own solution, of course. The other radical resolution is to say that sensory experience plays no role in our understanding of the concepts of the things and properties around. This is the standpoint of contemporary epiphenomenalism about conscious experience.

In this essay I will try to set out a straightforward resolution of Berkeley’s Puzzle: that is, one that accepts both (a) that our grasp of concepts of ordinary objects is grounded in experience of those objects, and (b) that we have the conception of mind-independent objects. To set out the straightforward resolution of Berkeley’s Puzzle we have to look a lot more critically at the idea of our grasp of concepts of ordinary objects being ‘grounded in’ experience of those objects; and we have to look a lot more critically at the idea of a conception of objects as ‘mind-independent.’ The most important single idea we have to have in play, though, to give a straightforward resolution of the Puzzle, is the conception of sensory experience as a relation between the perceiver and the scene observed. On a Relational View of Experience, sensory experience is characterized as a three-place relation holding between:

(i) the observer,
(ii) the point of view from which the scene is observed, and
(iii) the scene observed.

This three-place relation is “x is experiencing y from point of view z”, and as we shall discuss, it may itself be adverbially modified. The key point about any such view is that the scene observed cannot be eliminated from a description of the sensory experience. To be having a perceptual experience is, characteristically, to be experiencing a particular scene from a particular point of view; characterizing the qualitative character of the experience involves characterizing the qualitative characters of the objects and properties in the scene observed. The qualitative character of the sensory experience is constituted by the qualitative characters of the objects and properties in the scene observed. These are, as I have said, not the only components to a characterization of the experience; we have also to describe the point of view from which the scene is observed, and comment on any adverbial modifications of the relation of experience.

With this conception of sensory experience in hand, it is possible to resolve Berkeley’s Puzzle straightforwardly: that is, to acknowledge both the explanatory role of experience in our conception of the things around us, and our grasp of concepts of mind-independent objects. We first need to say something more about what it means for sensory experience to have a role in explaining our grasp of the conception of objects as mind-independent. Then we need to say something about how sensory experience, conceived as on the relational view, could play this role.

2. Ways of Thinking That Reflect Objectivity

There are two dimensions to grasp of a term referring to an ordinary physical object. There is, first, one’s grasp of the characteristic patterns of inference to which the term is subject. There is, second, one’s grasp of the justification for the use of those patterns of inference in connection with the term. That grasp of the semantic justification for the pattern of use of the term is what provides the most basic role for perceptual experience in our grasp of concepts relating to the medium-sized world.
There are characteristic ways in which we think and reason about ordinary physical objects that reflect the mind-independence we usually take physical things to have. We can think of the mind-independence of ordinary concrete objects as showing up in three different ways:

(1) It is the same object that we encounter over time. The object continues in existence even when it is not being observed. Even when the object is observed continuously over a period of time, there may be changes in the observation of it that do not reflect any change in the object, or any change of the object.

(2) It is the same object we encounter in different sensory modalities. You can see and touch one and the same thing, stick it in your mouth, and so on. There is one and the same thing here, through variation in the way it is being experienced by you.

(3) Different people can encounter one and the same object. The subjective experiences they have of the object are different to one another, but it is one and the same thing. So here there is variation both in who is having the experience and in what kind of experience it is, consistently with it being one concrete thing that is experienced.

It is fairly easy to see that Berkeley conceives of objects as mind-dependent. On his view, the identity of an object just is the identity of a perception, and two people cannot share the same perception, just as the same perception cannot be encountered in different sense-modalities or at different times. Yet obviously we do ordinarily talk in terms of one and the same object being encountered by different people, through different sense-modalities and at different times. How is it that we have the conception of it being one and the same object here? How is it that we can conceive of the unity of an object without conceiving of it in terms of its relation to a mind? The suggestion that has been most considered by theorists working on the subject is that our conception of the unity of an ordinary concrete object is the conception of a causal unity. To have the conception of concrete objects as mind-independent is to have some understanding of their identity-conditions; and those identity-conditions have to do with the causal dependence of the way the object is later on the way the object was earlier. When we characterize the identity of the object in this way, we do not need to bring in its relation to a mind. That is how we have the conception of the object as mind-independent.

When I put it like this, however, it may look as though we could give a comprehensive characterization of how one has the conception of objects as mind-independent without bringing in one’s experience of objects at all. For you could give a theoretical characterization of the causal dependencies here, and how the subject grasps them, without mentioning the subject’s experience of objects. So this way of explaining what it is to have the conception of objects as mind-independent may seem to lose any role for the subject’s experience of objects. All our understanding of objects as causal unities could be, as it were, theoretical, rather than being grounded in sensory experience.

To see just where experience comes in, notice that in general, we think of causal connections in two ways. First, we think in terms of causation as a matter of there being a counterfactual connection between two variables, Y and X. For Y to be counterfactually dependent on X is for it to be the case that the value of Y would have been different, had the value of X been different. The kinds of variables that matter for questions of the identity of objects are variables such as “whether saw Y cut through the wood easily at t2”, and “whether saw X had been oiled at t1”, where t2 is later than t1. If saw Y and saw X are different saws, then there will in general be no counterfactual dependence of the first variable on the second. That is, how easily saw Y is cutting through the wood would not have been any different whether or not you had oiled saw X. However, if the value of variables such as that first variable is counterfactually dependent on the value of the second variable, then that is the kind of causal connection that it takes for saw Y at t2 to be identical to saw X at t1. A grasp of this kind of counterfactual connection between variables relating to how the object is later and variables relating to how the object was earlier does not seem to exploit the subject’s experience of the object. (Of course, it is a familiar point that the counterfactuals here must not be “backtracking” counterfactuals. But I will not pursue that issue here.)

There is a second dimension to our ordinary thinking about causation. We do not think merely in terms of counterfactual connections between variables; we think in terms of mechanisms by means of which the counterfactual connections exist. Suppose we consider any finding established by a randomized controlled trial, such as a role for vitamin E in the prevention of heart disease. The trial might establish that interventions on the level of vitamin E in a patient will make a difference to the risk of heart disease. But there is a further, compelling question: What is the mechanism by which vitamin E affects the risk of heart disease? The answer: “There is no mechanism” would seem barely intelligible to most researchers. The truth of the counterfactuals is one thing, and the mechanism implicated in the causation here is something else. The notion of a mechanism here is notoriously difficult to characterize, and I am not going to attempt a general characterization of it, beyond remarking on its importance. For us the key point is that a prototypical case of a mechanism is provided by our ordinary conception of an ordinary medium sized object transmitting causal influence from place to place. This is one of the prototypical examples of a mechanism, and it provides us with one of the basic pictures of what a mechanism is that we bring to more difficult cases.

Let me give a simple example. Suppose I have two knives, knife A and knife B., and I have a knife-sharpener. Knife B is over at the cutting-board, with tomatoes ready to be chopped. Knife A is at the other end of the room, beside the knife-sharpener. Here are two cases:

Case 1. I sharpen knife A, and put it down beside the sharpener. I go to the other side of the room and use knife B to chop tomatoes. The chopping goes faster and better than it would have done if I had not sharpened knife A. Let us suppose that this is established by extensive experiment.

Case 2. I sharpen knife A. I take it over to the chopping board and use it to chop tomatoes. The chopping goes faster and better than it would have done if I had not sharpened the knife.

Case 1 is an unusual and puzzling case. How does it happen that the intervention at the sharpener is making a difference to what happens over at the cutting board? The natural question is, “What is the mechanism?”. Surely this must be some kind of conjuring trick. There must be some kind of machinery linking the two places, at the sharpener and at the cutting board, that we haven’t yet described. Case 2 seems quite different. What has happened at the sharpener is making a difference to what happens at the cutting board. But there is no puzzle about how this is happening. It is not that there is no mechanism here for transmitting causal influence from place to place. There is a mechanism. It is the movement of knife A. The movement of knife A from one place to another is the mechanism by which causal influence has been transmitted from one place to the other. This point is evidently quite general. The movements of concrete objects from place to place are the mechanisms by which causal influence is transmitted from place to place. This point is so familiar that it is easy to miss. But once you point it out, it is obviously basic to our thinking about causation in the immediate environment. That is what I mean by saying that concrete objects here are prototypical mechanisms by which causal influence is transmitted.

You might say that this is just a matter of what we are familiar with as opposed to what we are not familiar with. We are not familiar with Case 1 interventions whereas we are familiar with Case 2 interventions. But that response understates the contrast. Suppose Case 1 interventions did become commonplace. We would still look for a mechanism explaining why they happened. Similarly. we look for mechanisms explaining why light switches turn on lights, no matter how familiar we are with the phenomenon. It is not that there is some a priori guarantee that we will always find what we are looking for. It is just there is some looking to be done. In a Case 2 intervention, however, it is not just that the thing is familiar. The mechanism linking intervention and upshot is palpably there. Ordinary perception confronts us with the mechanism.
One reason it is easy to miss this point is that when we are looking for mechanisms we are often looking for something hidden. For example, it might take a lot of work to find the mechanism linking vitamin E with heart disease. But in this case the mechanism in play is perfectly obvious. You might point out that there is a hidden structure here, the molecular composition of the knife being transported from place to place. But this is not an alternative to thinking of the movement of the knife itself as the mechanism for the transmission of influence. All it provides us with is an equation: the movement of the knife is the movement of a collection of molecules. That identity may be true, but it is still the movement of the knife from place to place that is the mechanism for the transmission of causal influence.

To put this all together. A grasp of the identity of knife A over time may consist partly in a grasp of counterfactuals saying that what happened with the thing later would not have happened without what happened with the thing earlier. You know that an intervention on A earlier will make a difference to how it is later. But a grasp of the identity of A provides more than this. Your grasp of the identity of A means that you know the mechanism linking what happens at the earlier place to what happens at the later place. In contrast, you could know that an intervention on A is going to make a difference to how B is later. But you might still not know the mechanism involved. Knowledge of the mechanism involved is something over and above knowledge of the counterfactual. This understanding of the physical object as mechanism is not a matter merely of knowing, explicitly or implicitly, various counterfactuals about what would happen under interventions. In the absence of experience of objects, it is difficult to see how we could have such a conception at all. It is our sensory confrontation with the categorical object itself that provides us with our grasp of that which makes the counterfactuals true.
Suppose we do have this conception of physical objects as the prototypical mechanisms for transmitting causal influence from place to place. This is a conception of the concrete object as mind-independent: you think of the mechanism as having a unity that has nothing to do with its relation to any mind. A grasp of this idea cannot be explained as a matter of, for example, understanding in practice that your actions on the object earlier will make a difference to how it is later. You could grasp that there was this fact about the upshot of your actions without having any idea what mechanisms are involved. Our ordinary conception of a concrete physical object does provide us with an understanding of a mechanism, which must go beyond a mere practical grasp of intervention counterfactuals. So in virtue of what do we have this conception of the object as a mind-independent unity? The great merit of the Relational View of Experience is that it lets us understand how experience of the object could provide us with this conception.

3. The Relational View of Experience

We reason in ways that reflect our conception of objects as mind-independent. We engage in reasoning to establish object-identity that proceeds by arguing that the later object is the way it is only because the earlier object was the way it was; therefore, they are identical. Suppose, for example, that you visit your old schoolroom and find your initials still carved on a desk. You have the right to take it that it is the same desk; the initials would not be there now if it was not for your industry all those years ago. More fundamentally, we take ourselves to have a conception of “the desk itself,” on which the concrete object functions as the mechanism by which the influence of your earlier work is transmitted over time to the present. These ways of thinking reflect a conception of the object as mind-independent. Our question is whether we have any justification for reasoning in this way, and in particular, whether perceptual experience could provide us with any justification for reasoning in this way. Berkeley’s argument was that perceptual experience provides us with no such justification. Since it is only perceptual experience that can justify our use of particular patterns of reasoning about our surroundings, it follows that these patterns of reasoning are not legitimate and should be abandoned.
An alternative reaction, as I have said, is provided by contemporary epiphenomenalism, according to which perceptual experience plays no role in our knowledge of our surroundings, that is, no role in our grasp of concepts of the objects and properties around us. Someone who holds this view may say that there is no justification to be given of those ordinary ways of reasoning. This is just the way of our people, to talk and think in terms of mind-independent objects. Here we reach bedrock and the spade is turned.
Neither of these reactions seems particularly compelling. On the one hand, it is hard to believe that perceptual experience licenses us only in thinking and reasoning in terms of objects that are mind-dependent: transient, modality-specific, and perceiver-specific. The view that we can think only in terms of mind-dependent objects is simply impossible to believe, when ordinary perceptual experience seems to provide us at every moment with crowds of mind-independent things. On the other hand, the idea that we actually have no justification for thinking in this way—the idea that perceptual experience cannot underwrite these styles of reasoning—is equally hard to believe. These patterns of reasoning are not simply plucked out of the air, or imposed by us on a malleable reality—perceptual experience confronts us with the very mind-independent things whose being there validates these patterns of inference. That is what makes Berkeley’s Puzzle inescapable. We have to explain how it can be that sensory experience can be validating our reasoning about mind-independent material things.

In particular, let us focus on our ordinary conception of material objects as the mechanisms by which causal influence is transmitted from place to place and time to time. If we think of perceptual experience as a relation between the perceiver and the ordinary, mind-independent objects in the environment, then we can immediately see how perceptual experience could play a role in our having this conception. The key point is this:

On the relational view, the qualitative character of the experience is the qualitative character of the object itself.

As I brought out earlier, and I will return to the point in a moment, there is more to be said than this about the qualitative character of experience, on a relational view. There is the point of view from which the scene is being observed, and there may be adverbial modification of the type of experience in question. But the key point is that once these other parameters are set, on the relational view, the qualitative character of the experience is then constituted by the qualitative character of the object. That is what is distinctive of the relational view of experience.

It is because of this that the relational view of experience can explain how it is that we have the conception of the categorical concrete object as the underlying mechanism by which causal influence is transmitted. The concrete object, such as the knife we discussed earlier, is not a hypothesized cause; it is not known merely as a locus of counterfactual causal connections. Sensory experience brings the thing itself into view, the categorical basis of the counterfactual connections. Consider, in contrast, analyses of perceptual experience on which there is a distinction to be drawn between the qualitative character of the experience and the qualitative character of the object. That means that, on these analyses the experience itself does not bring the qualitative character of the object into the subjective life of the perceiver. The qualitative character of the experience is at best an effect of the qualitative character of the object perceived, on these analyses. So these analyses cannot explain how it is that a sensory encounter with the object can be an element in one’s grasp of the idea of the object as the categorical basis for the counterfactual connections between the way the thing is at one time and place and the way the thing is at another time and place.

The relational view of experience can similarly justify the patterns of inference that reflect our conception of objects as mind-independent. Suppose, for instance, that you consider some exercise of reasoning that uses the conception of sameness of object over time. Then, I am suggesting, what validates the correctness of the use of that pattern of reasoning is that the object is identified by your experience of it; and your experience of it is experience of a mind-independent object that is indeed the same over time, despite variation in other aspects of your sensory experience of it, or through gaps in your sensory experience of it. Similarly, in the cross-modal case, what justifies you in using a pattern of reasoning that uses the conception of sameness of object across sensory modality is the fact that your perceptual experiences in different sensory modalities have related you to one and the same object. And in the interpersonal case, what justifies you in reasoning in a way that exploits the sameness of the object that you and the other person have encountered is that you have both been experientially related to one and the same object. Of course, it can happen that you think you have justification for using a particular pattern of inference when you do not. Despite it seeming for all the world as though you have encountered the same object as one time as at another, there may in fact have been a switch. You can take yourself to seeing and feeling the same object when actually the thing you are seeing is not the same as the thing you are feeling. You can take yourself to be encountering the same object as someone else is encountering when in fact you are related to different things. In such cases your perceptual experience is not what you take it to be. Your perceptual experience does not validate the patterns of inference that you suppose it to. Moreover, even in the cases in which it is in fact the same object that is being encountered, across times or modalities or persons, that mere sameness of object is not enough for perceptual experience to validate the correctness of inferences that exploit the sameness of the object. As we saw already, in setting out the relational view of experience, it would not be right to think of perceptual experience as merely a two-place relation between the perceiver and an object. You experience scenes from a particular point of view, and there may be adverbial modifications of the relation of experience. So we have to say that for one’s perceptual experiences to justify the use of patterns of inference that exploit the sameness of the object, it must be not just that those experiences do in fact relate one to a single object, they all have to relate one to that object in suitably related ways.

Our understanding of concepts of the mind-independent world is not provided simply by our having perceptual experiences into which, as it were, mind-independent objects float; it is provided by our attention to the mind-independent objects we experience. Sometimes we have the right to take it that we are experiencing one and the same mind-independent object again, as we do when trading on identity, without any auxiliary argument to establish the truth of an identity proposition. And sometimes we do need some auxiliary argument before we can take it that we have one and the same thing again. To understand the distinction between those cases we have to look at the structure of conscious attention. The fundamental point to bear in mind here is that attention is itself a relation between the observer and the scene viewed. And in characterizing this relation, we have to describe the causal structure relating the objects and properties that the subject is experiencing.

To illustrate, let me take a particularly simple example where it is particularly clear what is going on. Suppose we consider the kind of diagram used for tests of color vision, in which a target object, say the number 5, is constituted of blobs of various sizes and luminances, on a dappled or brindled field of blobs similarly varying in size and luminance, so that the only systemic way in which the 5 is differentiated from its background is by hue. If someone can see the 5, then they do have visual experience of color; that is the assumption on which the color test is used (the Ishihara plates).
When someone can see the 5, the color and location of the object are the properties of it that are causally responsible for the object being singled out at all. That color has to be used to single out the object in visual experience is the whole point of the test; there is nothing other than hue that could causally differentiate the thing from its background. Of course location is being used, as become evident if you consider a variant on the Ishihara plate that has two figure 5s, of the same hue as one another, at different places on the same dappled or brindled background. All that differentiates the two 5s is their location; but you have absolutely no difficulty in focusing on one rather than another of the two 5s. So there is a contrast here between hue and location on the one hand, and shape on the other. Hue and location are causally involved in allowing you to see the object against its background. But the shape of the thing—whether it is a 5 or a 3 or something else—may be playing no such causal role. You would still be singling out the object on the basis of its hue and location whatever shape it had.

The fact that color and location are playing this causal role in your visual experience does not mean, however, that you have any knowledge of the color or location of the figure. The capacity to see the object is obviously in principle independent of the ability to give a verbal report of its color, for example, and independent of the ability to engage in color induction (“the red berries I have had all made me sick, so all red berries will probably make me sick”), or the ability even simply to match objects of different colors or to arrange them by color (for example, from darkest green to lightest green). And in fact children seem to have color vision in place as early as three or four months, an age that seems quite likely to predate any explicit knowledge of the colors of things. The function of color vision in animals is most importantly to allow them to see objects against their backgrounds. If color vision serves that function, the color may be of no further interest, and certainly not of interest for its own sake. So the causal role of a property in allowing you to see an object has to be differentiated from the causal role of a property in generating perceptual knowledge that the object has that property.

So suppose you can see the figure 5 against its background, and you form the judgment, “That figure is a 5”. The simplest form of the relational analysis describes the situation by saying that you are related by the two-place relation of consciousness to the figure 5, and to the properties of color, location, and shape. But what I have just been arguing is that we can give a more informative analysis of the situation by distinguishing between the kinds of causal roles played by the properties of color and location on the one hand, and shape on the other. We could find state this distinction by talking about the “point of view” from which the scene is being observed, or we could put it in terms of adverbial modification of the kind of experience of the scene one is having. And this kind of causal structure in one’s observation of the scene is what underpins one’s selection of a figure as the target of one’s attention, and a particular dimension—shape or orientation or whatever—of the object as the aspect of it about which the experience can generate knowledge.

Of course, this is a simple example, but the point I am making here seems to me quite general. Whenever an object is differentiated from its background, there will have to be some account of the characteristics of the object that, together with its location, are causing the object to be experienced as figure on ground. This is the realm of classical Gestalt psychology, or more recently, work on perceptual organization. And there will have to be some account of the dimensions of the object to which one is attending, or in other words, which properties of the object are directly implicated in knowledge that the object has them.

When we are characterizing your perceptual experience of an object, then, it will not be enough simply to say which object you are experiencing; we will also have to say which properties are being used to differentiate the object as figure from ground in your experience (color and location in our above example; but there are endlessly many further possibilities). This is a matter of characterizing the attentional structure of your experience. Our concern was with the validation of reasoning that relates to the identity of the object across time, or across modality, or across person. What we can say, on the relational view, is that for perceptual experience to underwrite such reasoning, it is not enough that the perceptual experience should relate to a single object throughout. We also have to take into account the way in which the object is being singled out in experience.

We can describe the way in which the object is being singled out by specifying the characteristics that are being used to differentiate the object as figure from its ground (color and location in our above example). Notice that this is not a matter of replacing the reference to the object in a description of the experience with a mention of these distinguishing characteristics. Rather, the experience is described as an experience of that very object, it being distinguished by those characteristics (such as color and location). Formally, we could incorporate those characteristics into a specification of the “point of view” from which the object is being observed, or we could think of them as part of a complex adverbial qualification of the way in which the object is being experienced.

In saying how perceptual experience validates the patterns of reasoning that we use concerning mind-independent objects, we have to reckon in the ways in which we single out the objects we perceive. These will play a role in determining when it is legitimate to trade on identity in one’s use of a term referring to the experienced object, and in providing the bases from which one can argue to establish the truth of identity statements relating to the experienced object. We will also have to reckon in the factors that allow us to keep track perceptually of a single object over a period of time, across modality and across speakers. Again, these factors will not replace the role of specification of the object itself in a characterization of the experience; rather they supplement the specification of the object, and provide a fuller description of the perceptual experience.

4. Other Minds and the Relational View of Experience

Thinking of sensory experience in terms of the relational view of experience is quite an alien exercise for many readers; so much so that it may be helpful if in this section I set out the bearing of this view on some more familiar ways of thinking of sensory experience. Our general question is whether phenomenal consciousness plays any role in our having a conception of the world around us, and if so, what that role might be. There are a number of aspects of perception that we might reasonably want to separate from the phenomenal or sensory awareness that I want to focus on. There is, for example, the mere fact that perception involves some kind of representation in the brain. In visual perception, for example, the visual system in the brain generates some kind of representation of one’s surroundings. In principle, this could happen whether or not there is sensory awareness of your surroundings. There are, for example, types of subliminal perception, in which we seem to have differential reactions to various stimuli, which seem to involve representation of them, without sensory awareness. Most strikingly, there is semantic activation without conscious identification. Presented with a word just below the threshold of awareness, you might respond in a way that seems dictated by the meaning of the word even though you are not aware of the presence of the word. In such cases it is hard to explain what is going on without assuming that you have some brain representation of the word itself, even though you have no awareness of it. Representation is one thing, and awareness is another. Sensory awareness is what happens when, in Nagel’s (2002) famous phrase, there is something it is like to perceive.

This notion of there being “something it is like” to perceive a thing needs, though, to be treated with some caution for our present purposes. The problem is that it is natural to give the notion an “internalist” twist that makes it difficult to achieve understanding of the epistemic role of consciousness. Let me explain first why the internalist twist can seem natural. Nagel introduced this notion in the context of imagining how the world is from a bat’s point of view. He remarked that one could know all there is to know about the physics of the bat and of the surroundings of the bat without knowing the one thing that seems most intriguing and elusive; what it is like to be a bat. Now when you put it like that, “being a bat” is something that seems to involve only the bat itself. So imagining what it is like to be a bat should involve only aspects of the world that are internal to the bat itself. But this is not way we usually understand the phrase, “what it is like”.

Suppose that your job is to design the stage set for a forthcoming performance. You have reached the point where you are on the stage surrounded by various pieces and a lot of machinery so you can manipulate your provisional set. However, the auditorium is being repaired so you can not sit in the stalls and view your set directly from there. You have to imagine what your set will be like from there. Will the red and gold you are using be a bit overwhelming, or will the distance have a subduing effect so that it merely seems sumptuously ornate? The set is a bit complicated. Will it seem cluttered, or is its spatial organization lucid enough that it will be evident what is going on? Will the actors be hidden from the spectators by the set, or will it frame them well? More briefly, what will it be like from, say, the rear stalls? That is the imaginative exercise you have to engage in. There is obviously no difficulty of principle about this, though some people will doubtless be better at it than others. Also, though, notice that this is a relational, or real-world exercise of the imagination. You are imagining what the set will be like from various vantage points in the auditorium. This is relational. It involves the physical stage set itself, the perceiver in the audience, and a relation between them.
For a philosopher who thinks about the conscious life in purely internalist terms, it will seem that this relational exercise misses the key thing. This is not a matter of imagining what the set will be like from various perspectives. To get onto the conscious life proper, you have to engage in a purely “internalist” exercise: imagining the mental life of the spectator in isolation from any of the surroundings, regarding the stage set itself as merely a causal prod to the production of inner experiences.
If you think of the project in this way, it is likely to strike you not merely as a skilled exercise that some people might be better at than others. The thing is likely to strike you as altogether impossible. After all, the spectrum of the individual in the stalls might be an inversion of yours, so that the sensations they have when viewing red and gold are more like the sensations you have when viewing green and grey. Or maybe they are nothing like any of your color sensations at all. Perhaps they are more like the sensations you have when listening to the tones of an organ. Or perhaps they are frankly alien, like nothing in your experience. How could you know? A stage-set designer struck by this thought would really have no way of proceeding. Perhaps God can know what the arbitrary individual in the stalls would experience, but it may be altogether beyond the ken of the humble set designer.

On a relational view of experience, there is no such “deep” inquiry into the sensory experience of the other to engage in. In the case of ordinary perceptual experience, there is no such thing as the “inner life” of the Other, considered in isolation from its relations to its surroundings. There is only the “surface” enquiry that set-designers and other visual artists and technicians routinely engage in. This simply does not require grappling with the unfathomable. This is the exercise of imagining, for example, what the set is like from the rear stalls. This is certainly a question about the conscious life of the perceiver; it is a question about what kind of sensory experience the audience will enjoy, and this impacts the kind of aesthetic experience the audience will have. But as I have said, it is a relational question. It is a question about the set itself, and the audience’s experiential relation to the set.

In the case of the bat, there is an analogue to the stage set designer’s practical question, but it seems, on a relational view, altogether unanswerable using only the resources of an ordinary human. The designer who wonders what the theater will be like from the point of view of, let us say, the bats hanging from the rafters, really may find that an understanding of this is beyond his powers. That is not because his question is an intrinsically unanswerable one about the internal mental life of another. It is rather that in this particular case, since it is a bat we are dealing with, many of the things that the bat experiences may not even be visible to the designer, and even if they are visible, the designer may have no idea which things they are. So relationally imagining the bat’s sensory experience may be impossible simply because one has no sense of which things the bat is experientially related to. In the case of the bat, it is very difficult for humans to imagine occupying the viewpoint of such a perceiver.

To say all this about our knowledge of another human’s sensory experience is not to deny that there is such a thing as depth in one’s understanding of another person’s mental life. Borges somewhere introduces one of his characters by saying, “Like all men, he was fathomless,” and few people would say that they do not recognize that sense of the endlessness of the probing it is possible to do into the inner life of another person, male or female. The present point is only that it is quite wrong to suppose that the right location for that sense of depth is in the characterization of sensory experience. If anything is a superficial fact about someone else’s mental life, it is the nature of their sensory experience.

I want now to give another way of indicating what I mean by “sensory experience.” This is to give a slight twist to Frank Jackson’s example of a person who lives in an entirely black-and-white environment, and learns all there is to know about color and color vision from black-and-white texts and diagrams. Jackson’s (1983) point was that this person, Mary, may have all the relevant physical information about her world, yet still learn something when she steps from her black-and-white room into the world of color. The slight twist I want to give to this poignant story is that whatever Mary learns, it is not in virtue simply of perception that she learns it; she learns in virtue of her sensory experience of her surroundings. There is a dispute in the literature on blindsight as to whether it is possible for blindsighted subjects reliably to make color discriminations on the basis of perception without awareness. For the purposes of our thought experiment, we can suppose that Mary is capable of discriminating between colors on the basis of perception without awareness. Even if this is true, it does not in any obvious way provide her with more information about her world than she had on the basis of the purely physical information given in her black-and-white texts. What she needs, to learn something, is sensory experience of the colors. This is not merely a matter of being able to make discriminations between colors on the basis of vision; we ordinarily have something more than that, the phenomenal blaze of colors themselves.

In terms of this example, we can see one of the key questions about the role of experience in our knowledge of the world. The question is whether what Mary learns on the basis of sensory experience of color is something about the nature of sensory experience itself. Since Jackson stated the example, it has always been taken that all that Mary can learn from her sensory experience of color is something about sensory experience of color. If you state this idea in full generality, it gives us the basis of Berkeley’s Puzzle. The idea is that Mary’s experience of color can provide a basis for the concept of the experience of color itself, and not much else. She could not have had any conception of what it is to experience scarlet before she stepped into the world of color, and afterwards she has learned what it is to experience color. More generally,

Sensory experience provides only for the concept of sensory experience itself.

You might argue that there is also another kind of “concept formation” that does not demand a sensory encounter with the phenomenon itself. You might argue that this other kind of concept formation demands only that you have been causally affected in one way or another by some distal stimulus; perhaps in this second sense that is all that is required for you to “have the concept” of the distal stimulus.

At this point we can already see something of the force of Berkeley’s Puzzle. You might already remark that it is hard to resist the sense that there is a robust notion of “knowing what the thing is” that you have for sensory experience, and that never applies to your knowledge of anything else at all. It is only a misleading compliment you pay yourself to say that you ever have the conception of anything else at all. And once you have reached this point, you are in the grip of Berkeley’s Puzzle. To have the concept of anything at all, including the concept of a mind-independent object, you have to have experience of the thing. But experience can provide only for the concept of experience itself. So we cannot, in any real sense, form the concept of a mind-independent object. We can form the concept only of experience itself.

My general point about Jackson’s Mary is that we can put the emphasis in a different place than it is usually put. Discussions of Jackson’s Mary usually put the key role of “sensory experience” as being what Mary learns about; they take it for granted that what Mary acquires on stepping from her room is some conception of sensory experience, and then they go on to argue about whether this fact is consistent with physicalism. My present point is that it is not obvious that what Mary learns about is sensory experience itself. Rather, our starting-point should be that it is in virtue of her sensory experience that Mary learns something. But it is not at all obvious that what she learns about is sensory experience itself. Rather, she may be learning about the color properties of the objects in her surroundings, even if she has no conception of mental states at all. The idea that sensory experience provides only for the concept of sensory experience itself is much more problematic than it is usually taken to be. Suppose we go back to the question: what does Mary learn when she steps from her black-and-white room into the blaze of color? Suppose that Mary does not so much as have the concept of a visual experience. It is sometimes suggested that people with autism have difficulty with the very concept of mind. Suppose we have a subject for whom something like this diagnosis is correct. It never occurs to her that other people have sensory experiences. It has never even occurred to her that she has sensory experiences herself. Or even if it has crossed her mind at some point in the past that others, and she herself, have sensory experiences, the phenomenon is one that she finds of no interest, and in practice she never directs her attention to questions about what sensory experiences people are having. So when this Mary steps out from her black-and-white room, whatever she learns about, she does not learn about sensory experience. She is interested in the characteristics of the concrete objects around her. She has, however, no interest in the psychological. She is interested only in the material characteristics of the physical things around her. Does our Mary learn anything when she steps out from the room? On the face of it she does. It will certainly seem to her that she does learn something. She will take it that she learns about a characteristic of the objects around her; a characteristic that does not have anything particularly to do with sensory experience except that she gets knowledge of it from the use of sensory experience.

One way to develop this point is to consider the case of someone who is a bit like Mary, only her case centers not on colors but on shapes. It is a bit harder to get this case stated. It is relatively easy to think of someone who occupies an entirely black-and-white environment. It is a bit harder to think of someone who occupies an environment devoid of shapes. The point is that someone who encounters only black-and-white objects does not seem to have the resources to define any other color concepts at all. But it is not quite clear how we could have a suitably non-spatial environment. On the face of it, if our subject encounters spatial characteristics at all she will have the resources from which to define spatial concepts; or at any rate, if something is lacking, it will not be something spatial that is lacking. But we can get the effect we want by again considering the direction of our subject’s attention. Suppose that our subject learns all about shape properties from purely textual or verbal information—no diagrams or photographs. And though she has ordinary spatial perception, she simply never attends selectively to the shapes of things. So it never occurs to her that the shape properties about which she has read so much are properties that are visibly possessed by the objects she perceives. What, if anything, does she learn when she finally makes that connection, when she finally grasps that the properties about which she has read are the same as the visible properties to which her attention has finally been drawn? On the face of it, our character—call her Bridget—will learn something. She will take herself to have learned something. You might say that what she has learned is something about mental states—she has learned something about what kind of visual experiences these properties produce. But we can suppose that Bridget has the same lack of interest in mental states generally, and visual experiences particularly, as our Mary. So whatever she learns, it is not something about visual experience. Bridget will nonetheless take herself to have learned something. She will have learned which properties those are that her texts concerned. On the basis of her sensory experience, she will have learned something about a physical characteristic of concrete objects. And notice that she would not have learnt this from a mere capacity to engage in blindsight-style reliable guessing as to which verbal shape-names apply to which objects in her vicinity. Sensory experience was really required for the learning here.

Let me give one last way of setting out the notion of “phenomenal awareness” or “sensory experience” that concerns me here. This is to appeal to Block’s (1997) distinction between access and phenomenal consciousness. The distinction here is between a notion of “access” consciousness that is defined in terms of the architecture of perceptual representations, and a notion that is not so defined. A representation is said to be “access” conscious if it is “poised for the rational control of thought and action”. I hope it is evident that our question does not have to do with the epistemic role of access consciousness. Our question is whether phenomenal consciousness makes a difference to one’s concepts of one’s surroundings. What is phenomenal consciousness? In giving his general characterization of the notion, Block relies on Nagel’s formulation: there is something it is like to be phenomenally conscious. So the comments made above apply here too—phenomenal consciousness may be a matter of how the world is grasped from a particular point of view, rather than a matter of purely internalist states. For example, the blindsight subject characteristically does not have the same internal functional architecture as the ordinarily sighted person; but seems to be missing something more, namely an ability to experience the objects and properties in the environment. What the blindsighter misses, in the blind field, is the kind of experiential contact with objects and their properties that could play a role in justifying the use of particular patterns of reasoning in connection with them. Of course, the blindsighted subject may formulate descriptions that uniquely specify objects in the blind field: “the object on my left”, “the thing the experimenter is manipulating”, and so on. But the blindsighted encounter with an object in the blind field is not enough to allow the subject to formulate the simple, perception-based demonstrative, “that thing”. The relationalist analysis applies to phenomenal consciousness. It says that the phenomenal consciousness involved in ordinary perception is a matter of the observer being experientially related to the scene observed, and that this is the basis of the observer’s knowledge of the objects and properties in question.

5. Sensations and Representations

The literature on sensory experience has tended to suppose that sensory experience must be characterized not in relational terms, but as a matter of having perceptual representations of one’s surroundings, or as a matter of having perceptual sensations. Neither of these views is particularly persuasive, taken on its own, so the literature largely consists of devotees of one side pointing out the problems with the other side, and in reaction, a massive and complex collection of compromise views, with some markedly radical positions from uncompromising sensationalists or uncompromising representationalists also being staked out. Berkeley’s Puzzle writes large the difficulties with these views; and indeed, the difficulty with the idea that sensationalism and representationalism are the only alternatives when it comes to the characterization of sensory experience. Suppose that we accept the dichotomy between sensation and representation in describing experience. What must sensory experience be, if it is to ground our conception of mind-independent objects? It is natural to think that sensory experience is really a matter of having sensations of one kind or another. But on the face of it, just having a sensation can not, in the first instance, provide you with information about anything except the sensation itself. You might, of course, form conjectures about the hypothesized causes and effects of the sensation. But a mere sensation in itself could not provide concepts of the objects and properties around you.

Alternatively, you might think that having a sensory experience is a matter of representing the world in a particular way. That seems to make it much more evident how sensory experience could, as it were, provide you with a window on the world. If your perception represents things as being thus-and-so, maybe that could explain how you know that things are thus-and-so. And if your experience represents the objects and properties around you, that can explain how your experience provides you with knowledge of the objects and properties around you, and so explain how you can have concepts of those objects and properties.

Suppose we ask how it is that sensory awareness, on the representationalist analysis, can provide you with an understanding of concepts: knowledge of the things and properties around you, knowledge of which objects and properties there are in your surroundings. The proposed answer is: this is a matter of having representations of those objects and properties, representations that figure in the content of your sensory awareness. Here we face the fundamental problem that you could have those very representations outside the context of conscious awareness. So if it is the representations that are providing the knowledge of objects and properties, we have as yet no explanation of how sensory experience might be doing any work in providing the knowledge. Moreover, it is in any case not obvious how having representations of colors or shapes, for example, could explain your knowledge of what colors and shapes are; representation of color or shape already seems to presuppose your grasp of concepts of color or shape. So it seems to get things the wrong way round, to say that your ability to represent colors and shapes explains how you know what colors and shapes are. You might argue that perceptual representations have a kind of representational content, “non-conceptual content”, that does not presuppose a grasp of concepts of ordinary objects and properties. But the more we stress that perceptual content is “lower-level” than conceptual content, the harder it becomes to see how a grasp of ordinary concepts could somehow be generated out of this kind of content. It is hard enough to define one set of concepts in terms of another set of concepts; there is no obvious way in which conceptual contents could be explained in terms of some quite different kind of content

On the one hand, it seems compelling to common sense that sensory experience has a fundamental epistemic role to play; sensory experience is our window on the world. On the other hand, it is not quite easy to see how that can be so. If we think of sensory experience as a matter of having sensations, then it is hard to see how sensory experience can be providing knowledge, in the first instance, of anything but those sensations themselves. If you think of sensory experience as a matter of representing how things are, then it is not obvious how awareness can have a fundamental epistemic role to play, since you can have representations without awareness; why would not representations without awareness provide just as good epistemic access to objects and properties? Moreover, the use of representations seems to presuppose, rather than to explain, the very knowledge of things and properties that we are trying to explain. The natural thought is that we should combine the representationalist and sensationalist approaches. Perhaps we can say that our representations of the mind-independent world are grounded in sensational aspects of perception. But now we face Berkeley’s argument. Put in terms of representation and sensation, his argument is this:

(1) There is a distinction between representational and sensational aspects of perceptual experience.
(2) All the representations we can form have the contents they do in virtue of their connections to the sensational aspects of experience.
(3) Representations that derive their meanings from their connections to sensation can only relate to how things are with sensations.

It seems to follow from this that we cannot form representations of a mind-independent reality.

It seems to me that what is problematic is the idea that the non-representational aspects of experience are to be thought of as sensations. As I have said, it seems to me arguable that the non-representational aspects of experience are not constituted by sensations. They are, rather, constituted by the experiential relations that the subject, occupying a particular point of view, stands in to the objects and properties in the environment. When we restate Berkeley’s assumptions in these terms, we have:

(i) There is a distinction between representational and relational aspects of perceptual experience.
(ii) All the representations we can form have the contents they do in virtue of their connections to the non-representational, relational aspects of experience.
(iii) Representations that derive their meanings form their connections to non-representational, relational aspects of experience may relate to how things are in an objective, mind-independent world.

The premise that would now be regarded as most obviously problematic in Berkeley’s argument is his (2). The argument for (2), though, is that only an approach that connects representational content to perceptual sensation can do justice to the sense in which representational content enters into the subjective life of the perceiver. Recasting the premise as (ii) aims to recognize this point.

Accepting premise (ii), you can still find a place for causal considerations in a theory of content. For causal considerations will matter for the existence of the relation “perceives” between subject and object, or subject and property; it is not to be supposed that what one is experiencing is a matter independent of some underlying level of physical and causal facts. Causal considerations matter because they have a constitutive role in determining what one is experiencing, and in consequence of that they have an impact on what one can represent or think about.

This gives us a way of recognizing what is still controversial and correct in Berkeley’s argument, while also acknowledging that our conscious experiences can ground thought about an objective, mind-independent world.

ACKNOWLEDGEMENT

This paper is excerpted from my contribution to a book jointly authored with Quassim Cassam, Berkeley’s Puzzle, that is in preparation. I am grateful to Quassim Cassam for comments on earlier drafts.

REFERENCES

Berkeley, G. 1734/1975. A Treatise Concerning the Principles of Human Knowledge. Reprinted in Berkeley: Philosophical Works, ed. M.R. Ayers (London: Everyman): pp. 62-127.

Block, N. 1995. On a Confusion About a Function of Consciousness. In The Nature of Consciousness, ed. N. Block, O. Flanagan, and G. Guzelder (Cambridge, MA: MIT Press).

Jackson, F. 2002. Epiphenomenal Qualia. In Philosophy of Mind: Classical and Contemporary Readings, ed. D. Chalmers (New York: Oxford University Press).

Nagel, T. 2002. What is it Like to be a Bat? In Philosophy of Mind: Classical and Contemporary Readings, ed. D. Chalmers (New York: Oxford University Press).

Estates have wings, and hang in Fortune’s pow’r
Loose on the point of ev’ry wav’ring Hour;
Ready, by force, or of their own accord,
By sale, at least by death, to change their Lord.

Alexander Pope, Imitation of Horace, 2nd Epistle of Second Book, 248-51.

1. Introduction
Most previous discussions of personal integrity have focused on the questions:
(i) What is the nature of integrity? and (ii) What is necessarily involved in the integrity of a person? Question (i) can seem intractable, asked straight out like that; the entry for “integrity” in the OED is certainly discouragingly various and contradictory. Question (ii) has seemed more hopeful, and it is generally expected that, in answering it, we can answer (i).

Discussions of (ii) reveal almost as many conceptions of integrity as there are contributors to the debate. But faced with the leading writers, and allowing for permutations and sub-divisions with further contrastive possibilities, the following four options seem to capture the basic positions:
The Unity Criterion: Integrity is essentially a matter of being united in agency; being integrated, intact, whole, undivided, or harmonious.

Advocates of the Unity criterion include Christine Korsgaard, John Cottingham, James Griffin, Gabrielle Taylor and Valerie Tiberius (1).

The Authenticity Criterion: Integrity is essentially a matter of being wholehearted; being true to one’s (“ true”) self; identifying with one’s actions, or motivations, or projects.
Advocates of the Authenticity criterion include Bernard Williams, Harry Frankfurt and Martin Hollis (2).

The Constancy Criterion: Integrity is essentially a matter of remaining consistent, steadfast, resolute, singleminded devoted.

Advocates of the Constancy criterion include John Rawls, Alasdair Macintyre, Michael Slote and Lynne McFall (3).

The Incorruptibility Criterion: Integrity is essentially a matter of remaining pure, innocent, decent, or free from taint.

Advocates of the Incorruptibility criterion include Sissela Bok (4).

In arguing for competing conceptions, and trying to resolve the issues arising, discussions of what is necessarily involved in the integrity of a person have been curiously self-limiting. This is because they have been based around cases that restrict themselves to a single choice on a single occasion. For example, “Jim and the Indians” and “George and the Laboratory” restrict themselves to a single situation and a single decision: whether or not to shoot a person and whether or not to accept a job, respectively. In so doing, they make time figure in an incidental way only; it is merely that in the course of which the relevant persons have developed the character and attitudes relevant to making their choices. It would not matter to the way these cases are usually discussed if—somehow—the person acquired both instantaneously.

I want to discuss the possibility that—sometimes at least—we need to take temporal relations into account in determining when and whether persons act with integrity. It may be, for example, that whether or not a person is to be considered as acting with integrity depends essentially on what is the case, or on what is predicted to be the case, at two or more different times.

Parfit’s case of the ‘Russian Nobleman’ provides a useful device for investigating this issue, and particularly our tendency to be gripped by inconsistent and contradictory ideas about what integrity over time is and what is valuable about it. The case turns on relations between actual and predicted situations with different temporal locations:

The Nineteenth Century Russian. In several years, a young Russian will inherit vast estates. Because he has socialist ideals, he intends, now, to give the land to the peasants. But he knows that in time his ideals may fade. To guard against this possibility, he does two things. He first signs a legal document, which will automatically give away the land, and which can only be revoked by his wife’s consent. He then says to his wife, ‘Promise me that, if I ever change my mind, and ask you to revoke this document, you will not consent.’ He adds, ‘I regard my ideals as essential to me. If I lose these ideals, I want you to think that I cease to exist. I want you to regard your husband then, not as me, the man who asks you for this promise, but only as his corrupted later self. Promise me that you would not do what he asks.’ […] The young man’s ideals fade, and in middle age he asks his wife to revoke the document. Though she promised him to refuse, he declares that he now releases her from her commitment…[She believes] that she is not released (5).

Discussions of this case have tended not to frame it in the most perspicuous way. Parfit told the story, but then concentrated on the wife’s position and actions alone. Those who have taken up the story—Williams, Elster, Velleman, Korsgaard, Brink, Lenman, Holton—have concentrated instead on the husband’s position and actions alone. One idea, whose plausibility I hope to demonstrate, is that the relevant issues for either partner fail to come into focus unless the positions and actions of both are treated with equal centrality, equal attentiveness.

The story is particularly useful for our purposes because it is far from offering a paradigm of integrity-behaviour, one which most could be expected to agree about. There may be something worth admiring in the married couple’s actions, for example, but it is really not obvious and self-evident that they manifest integrity. Some would say they do; some would say they do not. In getting clear about the reasons for disagreement, we might begin to get clear about integrity over time.

2.Integrity Criteria and Time
Can this pair be regarded as acting with integrity [6] ?
Here are some arguments, for and against. They are not meant to offer an exhaustive representation of the options available, but to locate features that discriminate between different criteria of integrity in their application to cases essentially involving time.

(A) The Authenticity Criterion
Those who believe in the Authenticity Criterion may easily argue that the husband does act with integrity, both when young and when middle-aged.

(Ai) When young, he has socialist ideals that generate projects with which he identifies himself. He is so wholehearted about this that he not only contracts to give away his estates when they come to him in the future, but extracts a promise from his wife to ensure that, come what may, the contract is not revoked.

(Aii) When “middle-aged and cynical” [7], he lacks ideals; the self he now identifies with looks with scorn on the naiveté of his youth. But he may nevertheless continue to act with integrity if this is also a self to which he can be true—by striving against doing what he no longer believes in, for example, or at least by not pretending he has ideals when he lacks them. The youth recognizes in himself this durable capacity for identifying wholeheartedly with whatever he takes himself to be. The point of extracting the promise from his wife, after all, was precisely to prevent what he regarded as the evil effects of combining middle-age with authenticity.

We have less information about the wife—despite the fact that Parfit raised the case to discuss how her actions might be defended. But this gives us leeway that can be dialectically useful. Those who believe in any one of the four criteria can easily argue either way. This is chiefly because of two factors which the story leaves undetermined: the identity of the persons over time, and the reasons for the wife’s willingness to make her promise at one time and her desire to keep it at another.
(Aiii) Those who believe in the Authenticity criterion can argue either way because what matters to them—that a person act in a wholehearted manner, one that counts as true to themselves—may, but need not, apply to the wife. That one should identify with what one does and thinks does not tell us what one should think and do. Evidently it is open to the wife to think differently about this. (This creates scope for too many versions of the story to discuss adequately here. I shall stick with a couple of variants throughout what follows; they give us enough shape to work on, without drowning us in possibilities).

(Aiv) Suppose we regard the middle-aged husband and wife as the same persons as those who were party to the original promising, and suppose the wife made her promise in a wholehearted way, just because her youthful husband begged her to. We could argue that she acted with integrity then and continues to do so in refusing her middle-aged husband. This might be because what she now identifies herself with differs. Perhaps she keeps the promise because she has acquired the socialist ideals her husband sloughed off. This need not conflict with the integrity claim because consistency is not essential to this criterion. Nor need it conflict with the personal identity claim. (There is a two-way independence between identity and ‘identify with’: one can ‘identify with’ conflicting types of belief and action at different times and remain the same person; equally, people can ‘identify with’ the same types of belief and action at the same time and remain different persons.)

(Av) Suppose we think that the identity of persons over time is such that qualitative change (of a sort consistent with this story) might destroy numerical identity: that the person of the early husband (or wife) might cease to exist; that the resultant, middle-aged husband (or wife) might be a different person. (Suppose, for example, we think that the identity of persons just consists in physical and/or psychological continuity, and that, due to lack of such continuity, the middle-aged couple are different persons from those party to the original promising.) If this failure of continuity is partly due to the fact that neither of them is now wholehearted about anything, then we can conclude that the wife lacks integrity whatever she does. If this is because the wife is now wholehearted about something else, we could argue she is acting with integrity: in refusing to revoke the document, she is being true to the way she now conceives herself to be.

(B) The Constancy Criterion
Those who believe in the Constancy Criterion may easily argue that the husband does not act with integrity, either when young or when middle-aged—at least if his forecast about his recidivism turns out to be accurate and it really is a case of allowing his ideals to fade away, i.e. as opposed to exposing them as false and replacing them with new ones.

(Bi) What matters to Constancy theorists is that he fails to remain consistent or steadfast. He surrenders his socialist ideals instead of being sufficiently devoted to them to stay resolute. Indeed, they will consider the device by which he binds his future self as evidence of this lack of integrity. Even whilst governed by his early ideals, he was lacking in single-mindedness, and sufficiently aware of this that it was necessary to construct the binding device. He never was a person of integrity.

As regards the wife, those who believe in the Constancy criterion can argue either way because what matters to them—that a person remain consistent, resolute—may, but need not, apply to her. Evidently such steadfastness requires, amongst other things, honoring the promises to which one is bound. But it is another question whether the wife is bound by her promise in her husband’s middle-age.
(Bii) Those who think that the identity of persons over time is such that qualitative change (of a sort consistent with this story) might destroy numerical identity can deny that there is sufficient continuity to identify the middle-aged husband with the person to whom she made her promise. Thus the promise she made is not one from which her middle-aged husband could release her. So in keeping that promise, she would be constant, steadfast, and thus acting with integrity. (This is the view Parfit takes.)

(Biii) If we reject this view of the identity of persons, or insist that there is insufficient discontinuity between the youth and the middle-aged man to mark a change of persons, we could argue that the wife is not acting in a consistent or resolute way. Suppose she makes the promise, not because she has socialist ideals herself, but because she does what her husband begs her to do, whenever possible—say because she knows that when he begs it is not an everyday desire he wants satisfied, but a matter essential to him. Since the middle-aged man is the same as the youth, he has the power to release her from her promise, and he begs her to regard herself as so released. In agreeing to do what he begs at one time, because he begs it, and then refusing to do what he begs at another time, when it is evidently in her power to do what he asks, she fails to be constant, consistent, and hence lacks integrity.

(Biv) Suppose we alter (Biii) in this one respect: the wife makes the promise, not just because her youthful husband begs her to, but because she shares his ideals—both factors are necessary and jointly sufficient explanations for her willingness to promise. Then her middle-age husband’s request creates a conflict that those who believe in the Constancy criterion also face. They could argue that there is no way for her to be constant in her commitment to both her socialist ideals and doing what her husband begs—hence it is possible that there is no way for her to act with integrity.

(C) The Incorruptibility Criterion
Those who believe in the Incorruptibility Criterion may equally easily argue that the husband—either when young or when middle-aged—does not act with integrity (at least if his forecast about his recidivism turns out to be accurate).

(Ci) What matters to Incorruptibility theorists is that in middle age he turns cynical and to that extent fails to remain pure or innocent. He surrenders his ideals altogether, rather than convincing himself of their falsity and replacing them with new ones. He may have been decent and free from taint at one time, but if he never possessed the resources to remain so, and knew that he lacked them (his recourse to the binding device is evidence of this), then he never was a person of integrity.

As regards the wife, those who believe in the Incorruptibility criterion can argue either way because what matters to them—that a person remain pure, innocent, decent, free from taint—may, but need not, apply to her. There is a significant constraint here: that innocence cannot be recovered once lost. If the reasons for which the person who first made the promise were corrupt, then that person cannot be acting with integrity whether or not they later decide to keep it.

(Cii) Suppose the wife is not corrupt at the point her middle-aged husband asks her to give up her promise. Nevertheless, she might refuse out of a desire to drive him to despair and suicide. We could then argue that the act is corrupt and blocks the possibility of her acting in the future with integrity.

(Ciii) Suppose the wife is corrupt at the point of making the promise but is not the same person when asked to renounce it. She may latterly have taken on socialist ideals, for example—if renouncing such ideals can create a new person, then presumably acquiring them can also. If her reasons are decent, we can argue that she is then acting with integrity.

(D) The Unity Criterion
As regards the husband, those who believe in the Unity Criterion can argue either way. This is because time can be made to figure differently in their appreciation. What matters to them is that a person of integrity manifests unity of agency, integration, or intactness. It is another question whether being integrated requires remaining so—for some appreciable period (there are, presumably, no fixed criteria for this).

(Di) Those who do not make integration depend on remaining integrated focus on whether a person can be considered united and intact in their action at a particular time. It is plausible to suppose that, on these grounds, both the young man and the middle-aged man act with integrity. There need be nothing disunited about their exercises of agency on each occasion. Evidently, there is considerable disunity between the young and middle-aged husband. But since what they believe and do here is separated significantly in time, it is irrelevant that what they believe and do differs markedly.
(Dii) Those who do make integration depend on remaining integrated for some appreciable period could regard both as acting with integrity. But this might require holding a revised view of the identity of persons. For suppose we think that the identity of persons over time is such that qualitative change (of a sort consistent with this story) might destroy numerical identity, and we agree with the young man that there is insufficient continuity to identify the person he is with the person of the middle-aged man. Then we cannot regard the differences between the two as undermining the integration of either one. The young man manifests unity of agency because he remains integrated so long as that person exists; likewise the middle-aged man. In other words, we would have a reason that is not ad hoc to deny that the ‘appreciable time’ over which the young husband would have to remain integrated to count as satisfying this interpretation of the Unity criterion must extend to that point in the future when he inherits the estates and must decide what to do.

(Diii) If, on the other hand, we reject this view of the identity of persons, or retain it while insisting that the discontinuity between the attitudes of the youth and the middle-aged man is not sufficient to mark the end of one person and the beginning of another, we can argue that there is a lack of integrity here. This might be the view most Unity theorists would take.
(Div) A weaker version of this view distinguishes the youth from the middle-aged man. The youth’s adherence to his ideals enables him to manifest unity of action for a sufficient time to count as remaining integrated. At some point, he loses intactness and with it integrity, turning into the middle-aged man.

(Dv) A strong version of this view denies that either the youth or the middle-aged man manifest integrity. Certain pieces of evidence—such as the youth’s dependence on a device to bind himself in the future—reveal that he never was intact; not even when he adhered to his ideals.

As regards the wife, those who believe in the Unity criterion can argue either way because what matters to them—that a person be united in agency, integrated, whole—may, but need not, apply to her.
(Dvi) Those who do not make integration depend on remaining integrated can argue straightforwardly that both the young and the middle-aged woman act with integrity. What they believe and do is significantly separated in time, so it is irrelevant that what they believe and do differs markedly.

(Dvii) Those who do make integration depend on remaining integrated for some appreciable period could regard both as acting with integrity. They would not need to take a particular view of the identity of persons to do so.

(Dviii) Suppose both the husband and the wife remain the same persons throughout. We could regard the unity of the wife as manifest in her willingness to keep her promise, to sustain the course of action she entered upon in making that promise, despite her husband’s recidivism.

(Dix) Suppose that the husband and the wife have become different persons. We could not then base the integratedness of the wife on the fact that she is the same person as the one who made the promise. But we could argue that her unity of agency is manifest in the fact that she bases her integratedness as the person she is now around a commitment to carrying out the projects to which another person (her ‘earlier self’) was committed.

3. The Unity Criterion
I have given a first description of several different views about integrity and time. These views make different claims about ordinary people and ordinary lives, even though we discussed them in terms of a fictional case.

Which—if any—of the criteria of integrity are correct? To answer this, we need to know more about what these criteria entail.

In the remainder of this paper, I shall look more closely at the Unity criterion. This is not because it is the dominating view of integrity. There is no such thing. (That is why the so-called ‘Integrity Objection’ seems intractable.) It is simply because it seems to me the most interesting view to examine. As we have just seen, the Unity criterion appears—at first glance, anyway—to be the most “open” of the alternatives. It is natural enough for Authenticity theorists to argue that the husband does act with integrity, and equally natural for Constancy and Incorruptibility theorists to argue that he does not. Not so for Unity theorists; it is unclear what, if anything, they would find it ‘most natural’ to say about the husband. Because the Unity criterion makes the case more troublesome, it is more interesting to consider it. It may also be more fruitful.

I have suggested that it would not be implausible for a Unity theorist to regard the husband as acting with integrity. This is so even if such a theorist took the view that to be integrated and united in agency requires remaining so for some appreciable period.

Korsgaard endorses the Unity criterion.8 But she argues in the diametrically opposed way (9). Call the point at which the husband decides to ask his wife for her promise Request Time. On her view:
(Main Conclusion) The youthful husband could not be regarded as acting with integrity at Request Time.

This is so even if we deny that to be integrated and united in agency requires remaining so for some appreciable period. For the youthful husband is, even at Request Time, “a mere heap of unrelated impulses,” and thus “in a condition of war with himself”—indeed, more precisely, in a condition of “civil war” (10).
Before examining this view, I should say at once that my intention is to keep the discussion in bounds and focused on the issue which immediately interests us: integrity over time. In effect, this means conceding to Korsgaard aspects of her position which lead away from that focus—and, in particular, granting certain controversial Kant-inspired claims about marriage and reasons for action which would, on other occasions, call for discussion.

Korsgaard offers two separable arguments for her main conclusion. The first argument aims to show that:

(I) The youthful husband’s attitude to his wife is such as to make acting with integrity impossible for him at Request Time. Call the argument for (I) the Marriage Disunity Argument. The second argument aims to show that

(II) The youthful husband’s attitude to his own future self is such as to make acting with integrity impossible for him at Request Time. Call the argument for (II) the Life Disunity Argument.

If they are sound, Korsgaard’s two arguments show independently that, regardless of what subsequently occurs, the youthful husband could not be regarded as integrated and united in agency at Request Time; that it would be inconsistent with Parfit’s story to suppose he could be; and that since integrity is a matter of being integrated and united in agency, the youthful husband could not be regarded as acting with integrity at Request Time.

It may immediately seem that there must be two significant differences between these arguments: that the Marriage Disunity Argument concerns a different person and the present time, whereas the Life Disunity argument concerns the same person and a future time. But this might be misleading for two reasons.

Distinguishing the arguments in this way might make it seem that considerations of time are significant to the Life Disunity argument and not to the Marriage Disunity Argument. This would furnish grounds for an immediate objection: that the Marriage Disunity argument is irrelevant to the issue of integrity over time. But this would be false.

The Marriage Disunity Argument turns on the husband’s attitude to his wife as his wife—that is, as someone whose future he is bound to by the married state. It is precisely his attitude to that future which must make him disunited, on Korsgaard’s view. Conversely, the Life Disunity argument turns on the youth’s attitude to his future self, i.e. as someone whose past self he now is. It is precisely his attitude in the present to that self which must make him disunited, in Korsgaard’s view. Hence both arguments concern both present and future.

There is a second reason to be concerned about this way of distinguishing the arguments. It might make it seem that, for the Life Disunity argument to work, one would have to think that the youthful husband and the middle-aged husband are the same person. But a revisionist about the identity of persons could deny that they are the same person. Hence such a revisionist would be furnished with an immediate objection: the Life Disunity argument begs the question, assuming precisely what they deny. But this would be false.

It is true that Korsgaard herself does not hold the revisionist view that would imply that a qualitative change (of a sort consistent with this story) might destroy numerical identity. She regards the youthful and middle-aged husband as the same person (11). But the Life Disunity argument might still go through even if she did hold the revisionist view.

What matters for the Life Disunity argument is the youthful husband’s attitude towards himself at Request Time; it is here that he fails to act with integrity, regardless of what subsequently occurs. It is in part because he regards the future person who will inherit the estates as not himself that he is disunited at Request Time. Suppose the youth is correct in his forecast of deep qualitative change. And suppose there were another philosopher, in other respects like Korsgaard, but willing to agree that the youth and the middle-aged husband are not the same person—Revisionist Korsgaard. Then the Marriage and Life Disunity arguments would not differ in respect of persons: both would be concerned with the youth’s attitudes toward what he takes to be—and is—a different person. But Revisionist Korsgaard could still apply the Life Disunity argument. This is because the fact that the youth is correct in his forecast does not make him any the less disunited in his agency.

Indeed, Revisionist Korsgaard is free to diagnose the youth’s prophecy as self-fulfilling: that it is precisely because of the attitude he takes towards his future self that the self he turns into is indeed a different person. However, this would be an additional claim, unnecessary for her purposes. The essential claim is that it is because the youth takes the attitude he does towards his future self that he is disunited at Request Time. And here Korsgaard and Revisionist Korsgaard are as one.
Beneath Korsgaard’s own—scattered—formulations of her position, we can discern a common form to the Marriage Disunity argument and the Life Disunity argument. In Korsgaard’s view, At Request Time the youthful husband simultaneously both tries to commit himself to a certain course of action and does not try to commit himself to that same course of action.

(Given (a)) The husband is “at war” with himself; engaged in “civil war”; he could not be regarded as integrated or united in agency.

(Given the Unity criterion) If someone could not be regarded as integrated or united in agency, they could not be regarded as acting with integrity.

(Given (a-c)) The youthful husband could not be regarded as acting with integrity at Request Time —this is Korsgaard’s main conclusion.

Evidently what matters for each realization of this form of argument—the Marriage Disunity argument and the Life Disunity argument—is the support they give (a).

Premise (a) evidently involves the youthful husband in a contradiction. The most plausible way to explain it, perhaps, is to say that of part of him it is true that he tries to commit himself to a certain course of action, and of part of him it is true that he does not try to commit himself to that same action. This would imply that he is fundamentally disunited, of course, and in a strong sense. Not only is he in parts, but those parts are in strenuous conflict. And that is precisely how Korsgaard sees it, hence her advocacy of (b).

Two questions are worth raising. The first is whether the evidence of conflict in (a) is indeed strong enough, or of the right sort, to support (b). If one is to count as fundamentally disunited and engaged in “civil war,” it may be necessary that one be fully self-conscious and self-reflective about the fact that one is both trying and not trying to do something. The second question is whether the evidence of conflict in (a) is indeed strong enough to support (c) when combined with (b). It may be necessary, for example, that one’s self-conflict be at a very deep level if it is to count as undermining one’s capacity for integrity. Korsgaard herself does not address either issue, and I shall not pursue them further in what follows.

4. The Marriage Disunity Argument
The Marriage Disunity argument aims to support (a) by claiming that, at Request Time, the youthful husband both tries and does not try to commit himself in marriage to his wife. The argument proceeds as follows:

At Request Time, the youthful husband tries to commit himself in marriage to his wife.
Korsgaard assumes this. She takes it on faith, I think, that the youthful husband is sincere and, having entered into the married state, he acts in such a way as to make (1) true.

Korsgaard takes from Kant’s Lectures on Ethics and the Metaphysics of Morals the idea that
To commit himself in marriage to his wife means that the husband must try to “unify his will” with hers.

By “unify his will” she means that the husband must make decisions together with her, “deliberate together” with her, and “arrive at a shared decision.”12

Korsgaard then argues, for reasons we will investigate, that: At Request Time, the youthful husband does not try to “unify his will” with that of his wife.

(Given 2 and 3) At Request Time, the youthful husband does not try to commit himself in marriage to his wife.

And hence: (Given 1 and 4) At Request Time the youthful husband simultaneously both tries to commit himself to a certain course of action and does not try to commit himself to that same course of action—that is, premise (a) above.

On the face of it, (2) seems a weak requirement on committing oneself in marriage. For two people, A and B, may be described as “deliberating together” even though A has the dominating and subjugating voice throughout the discussion. And they may be described as arriving at a “shared decision,” even though A obtains all she wants without taking what B wants into account, in the sense that both A and B consider themselves bound by the decision, and equally responsible for putting it into practice, and for dealing with the consequences. If this is all that is required for “unity of will,” then even committees with the most tyrannical of chairs can claim to meet it. This would not be such a problem were it not for the danger in which it places Korsgaard’s argument. For if the requirement on “unifying one’s will” is as weak as she makes it, what plausible reason can she offer to deny that the youthful husband satisfies it (that is, (3))?

Korsgaard’s argument for (3) is terse, but I think we can find within it two separable arguments.

The first turns on the idea that the youthful husband is not in a position even to try to commit himself:

To try to “unify his will” with that of his wife requires that the husband be capable of committing himself to his wife.

Korsgaard seems to be dependent here on some version of the claim that genuinely to be counted as trying to do something entails that one must be—or at least take oneself to be—capable of doing it.
At Request Time, the youthful husband asks his wife to promise not to revoke his contract.
Hence (ii) reveals that at Request Time, the youthful husband is incapable of committing himself (to anyone or anything).

This may seem an extraordinary claim, and one not required by the story. But on Korsgaard’s interpretation, the wife’s promise is precisely a device to ensure that, though the husband is incapable of committing himself, a commitment is nevertheless made, and kept, namely by her. She asks, rhetorically, “So what is she supposed to think of his marriage vows?”(13)

(Given (i) and (iii)) At Request Time, the youthful husband cannot try to ‘unify his will’ with that of his wife.

(Given (v)) At Request Time, the youthful husband does not try to ‘unify his will’ with that of his wife—that is, (3).

Korsgaard’s second argument for (3) turns on the position in which the youthful husband puts his wife by asking for her promise.

At Request Time, the youthful husband requests her promise not to revoke the contract he has made when he inherits his estates.

If she does not comply with this request, given what it means to him, then it will be impossible for her to “unify her will” with her husband’s.

Here also Korsgaard seems to depend on a much stronger requirement on “unifying the will” than she herself offers. For it is evidently possible for her to “deliberate together” with him, and—if she is strong-willed enough—to arrive at a “shared decision,” even though that decision entails her not complying with his request. She may, for example, insist that the scheme be abandoned, and obtain her husband’s agreement, even though it costs him dear. Or, if she is more accommodating, they may arrive at some compromise.

If she does comply with his request, she faces a future in which she has to choose between her loyalty to her youthful husband and her loyalty to what has become of him.
(Given viii) If she does comply with his request, she faces a future in which whatever course she takes, she has been forced to will as an independent person.

Korsgaard seems to have the following thought in mind. Suppose the wife carries out the wishes of the youth. Still, she has had to decide to do this independently of him: it was as an independent person that she chose to favour his wishes before those of the middle-aged man. The same would apply if she had favoured the middle-aged man instead, or taken any other course of action.

(Given (ix)) If she does comply with his request, it will be impossible for her to “unify her will” with her husband’s.

Presumably Korsgaard does not think that, once the wife’s will is “unified” with that of her husband, there is no respect in which she can will in the future as an independent person. That would be crazy. It must be that there are decisions of certain kinds that she cannot make independently whilst being so united (examples she herself offers include decisions about the house and the car), and the resolution to comply with his request is one of them.

(Given (vii) and (x)) At Request Time, the youthful husband makes it impossible for his wife to “unify her will” with his.

To try to “unify his will” with that of his wife requires that the husband avoid making it impossible for his wife to “unify her will” with his.

(Given (xi) and (xii)) At Request Time, the youthful husband does not try to “unify his will” with that of his wife—i.e. (3).

To summarize: Korsgaard takes the view that, at Request Time, the youthful husband puts his wife in a position where, whatever she does, she cannot decide in union with him; hence he has not tried to “unify his will” with hers and has not tried to be committed in marriage to her—that is, (4). In conjunction with (1), (4) entails (5) which is an instance of (a): i.e. at Request Time the youthful husband simultaneously both tries to commit himself to a certain course of action and does not try to commit himself to that same course of action. And in conjunction with (b)-(c), (a) entails Korsgaard’s main conclusion: i.e. the youthful husband could not be regarded as acting with integrity at Request Time.

The Marriage Disunity argument may be faulted. Korsgaard thinks that the youthful husband’s recourse to his device—requesting the promise from his wife—is incompatible with an attempt to “unify his will” with that of his wife. In response, we might say that the youthful husband’s recourse to this device is not simply consistent with the attempt to “unify his will” with that of his wife, but actually required by it.

To see why, consider a comment made by Korsgaard. She notes that, by making this promise, the wife “is to hold him, by holding herself, to giving up the estates,” and she asks, rhetorically, “But if she can do this, why can’t he?”14 But this is a strange complaint. There are many important projects to which one cannot commit oneself without others being committed too. It may be easier for others to be so committed (just as the wife’s role in this scheme is easier, perhaps; it is not she, after all, who stands to inherit the estates. It is only by virtue of marriage that she stands a chance of benefiting from them). In such cases, we accept that those who have it easier may be able to play their role while those who have it harder cannot.

Korsgaard might want to say that it would be better if the youthful husband were stronger in himself and thus had no need of such a device. This may be so. But the fact that he is not “stronger in himself” does not in itself make him incapable of “uniting his will” with that of his wife in marriage. Indeed, it can strengthen that capacity. For the husband might say, for example, that the device is less a way of having himself bound to a certain course of action in the future than a way of constructing a “unity of will” with his wife in the present—a unity which, he hopes, will flourish through the future, based partly on their joint commitment to a common project of living their life in accord with socialist principles and without dependency on a future inheritance.

Korsgaard could reply that this is false to the story. It makes it seem as if the youthful husband’s primary purpose is to “unify his will” with his wife, and requesting this promise from her is just a means of achieving that. But as Parfit tells the story, the youthful husband’s primary purpose is evidently to deal with his recidivism. He does not invent or play upon his recidivism to create opportunities for union with his wife. He clearly wishes he was not a recidivist.

We can adapt the objection to meet this response. As Parfit tells the story, the husband’s primary purpose is indeed to deal with his recidivism. He obliges himself to act as his youthful self judges best. He could have picked another as the means by which this is achieved, writing them into the contract and extracting a similar promise from them. But he precisely chooses to incorporate his wife into the means by which his recidivism is checked. In so doing, he makes his recidivism into an opportunity to “unify his will” with his wife. He does not treat her merely as a means in making this request of her. He freely offers her the chance to help him, precisely because he wants to ensure that she continue to have her own voice and role in this matter which touches her, as well as him, so completely.

We can strengthen this objection. The youthful husband might ask how else he is to live out the “unity of will” which the Kantian view of marriage enjoins on him unless he involves his wife—via her promise—as a free voice in the device by which he binds his future self. He might continue: it is surely required of him by this Kantian view that his way of dealing with his recidivism is not to hide it from his wife, or attempt to disguise it in some way, but to incorporate her—to the extent the laws of inheritance make possible—as a partner in dealing with it.

Korsgaard can reply that there is another option: to live out what marriage under the Kantian conception requires by finding a middle course, one which both his present and future self can endorse. Lenman concurs: it would resolve matters if the youthful husband thought of a possible life that is “choiceworthy” from both perspectives and lived it.15 But here it is we who have recourse to the details of Parfit’s story. For it is at least consistent with that story to deny that this is an option for the youthful husband.

As Parfit describes him, the man is in a similar position to Luther at the Diet of Worms (at least according to the disreputable way philosophers recall history16): renouncing his estates matters to him not just so much, but in such a way, that he cannot do anything else but renounce them. Having reached his position on the issue, there are indefinitely many things he cannot now do. Every apparent alternative to renouncing his estates is actually—for him—unthinkable. This is a situation, in other words, where he would consider it correct to say, “There is nothing else to think on the matter.” His deliberative decision not to do anything else, reached on the basis of considerations that are totally decisive for him, just is the conclusion that he cannot do it. This is not because of a deficient capacity on his part, an inability to take another course because of some internal compulsiveness, for example, or because of some external force with which he does not identify. It is because he has, so he believes, such good reason to reject any other course. These rational considerations act on him like a force, and one that he cannot bring himself to overcome.17 Indeed, the fact that doing anything else is simply not an option for him might naturally be taken precisely as a mark of his integrity.

If this is plausible, then the Unity theorist with Kantian views appears to be in an awkward position. Someone in similar circumstances to the youthful husband might decide not to renounce their estates in order to preserve their unity. This person might thereby count as a person of integrity. But this person is not like the youthful husband, and in a way that does not reflect badly on the youthful husband. For the decision not to renounce his estates is not an option for him. And the fact that this is not an option for him is plausibly to be taken as a mark of his integrity. Yet, in being unable to live out what marriage under the Kantian conception requires by finding a middle course, one that his future self could also endorse, it seems that he cannot be a person of integrity. How could we make this conclusion seem defensible?

One line alone seems available to the Unity theorist, who insists that integrity is a matter of unity and integratedness. If the youthful husband is indeed to be a person of integrity, he must defend and promote that unity and integratedness. Hence, in this instance, he must choose the middle course. This is so even if it means doing what, at this time, he regards as unthinkable, as that which he cannot do.

But this line would be self-defeating, and in a way that Korsgaard herself would mark out as salient. For if the youthful husband were to promote unity over and against what he feels he has no option but to do, he will be setting himself ‘at war’ with himself, and thus lose whatever unity and integratedness he has. In short, if integrity is a matter of unity, it cannot be consistent with integrity to promote unity over and against what one feels one has no option but to do.

So this is no solution to the awkward position. We have some reason to regard the youthful husband as a person of integrity. But if we retain the Unity criterion, at least in the way Korsgaard defends it, we must deny this. Acting with integrity is impossible for him. Now this is, of course, precisely the conclusion Korsgaard draws. But it is not for the reasons she gives. Indeed, it is for reasons that may—perhaps should—make her uncomfortable too: namely, it is not possible for the youthful husband to remain a person of integrity and care so much and in such a way about something that one course of action and no other is an option for him.

There is a way to make the discomfort more pointed, given Korsgaard’s Kantian conception of marriage. Suppose that what some person A cares about so much and in such a way is his spouse, or the success of his marriage. Suppose further that there are circumstances in which this makes one course of action and no other an option for him. Is it really acceptable to propose that, this being so, it is impossible for him to act with integrity?

5. The Life Disunity Argument
Before addressing this question, we should examine the second argument Korsgaard offers in support of her main conclusion. The Life Disunity argument aims to support (a) by claiming that, at Request Time, the youthful husband both tries and does not try to commit his own future self to giving his inheritance away. The argument proceeds as follows:

At Request Time, the youthful husband tries to commit his own future self to giving his inheritance away.

In drawing up his contract and extracting the promise from his wife, he acts in such a way as to make (6) true.

Korsgaard (18) takes from Kant’s Lectures on Ethics and the Metaphysics of Morals the idea that
To commit his own future self to giving his inheritance away means that the husband must try to “will a law” for himself right now—i.e. at Request Time.

She explains what she means by “will a law”:

To try to “will a law” for oneself right now is to accord normative standing to the reasons for which one is acting right now.

She explains what she means by “according normative standing”: To accord normative standing to the reasons for which he is acting right now, the husband must accord normative standing to the reasons for which his future self will propose to act.

She then argues that: At Request Time, the youthful husband does not accord normative standing to that for which his future self will propose to act.

Putting these claims together, she argues that: (Given 7-10) At Request Time, the youthful husband does not try to commit his own future self to giving his inheritance away.

And hence draws the conclusion: (Given 6 and 11) At Request Time the youthful husband simultaneously both tries to commit himself to a certain course of action and does not try to commit himself to that same course of action—i.e. (a).

To summarize: Korsgaard takes the view that, at Request Time, the youthful husband puts himself into a particular position with regard to his future self, one which makes him incapable of trying to commit that future self to giving his inheritance away—that is, (11). In conjunction with (6), (11) entails (12), which is an instance of (a): i.e. at Request Time the youthful husband simultaneously both tries to commit himself to a certain course of action and does not try to commit himself to that same course of action. And in conjunction with (b)-(c), (a) entails Korsgaard’s main conclusion: the youthful husband could not be regarded as acting with integrity at Request Time.

The Life Disunity argument may be faulted. We might start with (9), the claim that the youthful husband must accord normative standing to the reasons for which his future self will propose to act if he is to accord normative standing to the reasons for which he is acting right now. We might agree but block Korsgaard’s inference. For we might deny that the future in which such reasons retain normative standing must include the point at which he inherits the estates. For it may be that, to count as such, reasons cannot have normative standing for an instant. But it does not follow that, to count as such, they could never lose that standing in the indeterminate future.

Korsgaard does not respond to this possibility. But she anticipates one way of making it seem legitimate for the youthful husband to deny normative standing to that for which his future self will act: namely, that he thinks his future self will then be acting out of weakness (for example, a lack of self-control) or out of irrationality (for example, a clouded judgement). She accepts that this would be a legitimate excuse. So she would endorse a qualification of (9):

9*. To accord normative standing to the reasons for which he is acting right now, the husband must accord normative standing to the reasons for which his future self will propose to act—unless he thinks his future self will be acting out of weakness or irrationality.

But Korsgaard also appeals to the fact that Parfit expressly says the case is not like that of Odysseus having himself bound to the mast when sailing past the Sirens.19 She takes this to mean that the husband does not think of his future self as acting out of weakness or irrationality.

It may seem strange for Parfit (or Korsgaard) to deny that the youthful husband is like Odysseus. For evidently the cases are alike in fundamental respects. In asking his sailors to bind him to the mast, Odysseus does as the youthful husband does in extracting his wife’s promise: he uses others to prevent his future self doing what his present self does not want him to do, something that he knows his future self is otherwise liable to do. Moreover, although both Odysseus and the husband put themselves under the control of others, that is only possible because they are in control of themselves—as the captain in the ship and the husband in the marriage, respectively. It is precisely because they are both in control in the general case that they are able to put themselves under control in this specific instance.

But Korsgaard could reply: the cases nevertheless do differ, and in the relevant respect. For Odysseus anticipates that his future self, if not controlled, will act out of weakness or irrationality. But the youthful husband assumes his future self will act out of a genuine allegiance to “different values” (Korsgaard’s phrase).20 Hence he has no legitimate excuse for denying normative standing to that for which his future self will act.

This may be so. But it seems to produce a contradiction in her argument. For if the youthful husband does assume his future self will act not out of weakness or irrationality but genuine allegiance to ‘different values’, then surely he is according normative standing to that for which his future self will propose to act. Granted, the reasons his future self will be acting on are not reasons with which his present self can identify; they are not reasons for which his present self would act. But they are reasons nevertheless, and to be accorded normative standing. This produces a contradiction, of course, because it is the negation of (10).

The point can be expressed more strongly. It is not just that the youthful husband need not deny that that for which his future self acts has normative standing. He must accept that it has such standing if he is to think of his future self as acting out of genuine allegiance to ‘different values’, rather than out of weakness or irrationality.

The upshot is a constructive dilemma. Either the youthful husband does regard as reasons the “different values” on which his future self will propose to act, or he does not. If Korsgaard thinks he does, then she must reject (10): the youthful husband thinks of his future self as acting for values that are reasons, just not reasons with which he would now identify. If she thinks he does not, then she must acknowledge that the qualification in (9*) offers the youthful husband a let-out. For then he has a legitimate excuse not to accord normative standing to what his future self will be acting on (i.e. at the time of acquiring the inheritance), while nevertheless according normative standing to that on which he is acting right now (i.e. at Request Time). (If this is the truth of it, then the youthful husband would be similar to Odysseus, of course; for Odysseus freely accords normative standing to that on which he is acting when ordering his men to bind him to the mast while denying such standing to what his future self will (try to) act on, when under the sway of the Sirens’ song.) Either way, the youthful husband may be regarded as according normative standing to what he acts on at Request Time. If this is correct, then there is no reason to deny that, at this point, he successfully “wills a law” for himself.

Korsgaard may reply that there is middle ground: the youthful husband can think of his future self as acting out of genuine allegiance to “different values” without thereby according normative standing to what he thereby acts out of. How she would make good on that claim I do not know. She says at one point that the youthful husband thinks of that for which his future self will propose to act not as “reasons” but as “facts to contend with, as tools and obstacles.”21

But the attempt at exclusive contrast is unpersuasive. Some of the most formidable facts we have to contend with are what others count as reasons—it is precisely their being accounted reasons that can make them so formidable. Similarly, the youthful husband may think of his future attitude as a tool or obstacle to be contended with or got around. But that is how it is for him at Request Time. It certainly does not preclude his thinking of this future attitude as providing reasons for his future self at the time of inheritance. Indeed, it is precisely because it will provide reasons, perhaps, that he strives so hard to prevent his future self from acting on that attitude. Obstacles to one course of action are the more formidable for being reasons for another course of action. In short, at Request Time, the youthful husband can think of his future attitude as both a reason (for his future self) and an obstacle (for his present self).

So Korsgaard would have difficulty in finding middle ground. But suppose she can. Unfortunately, that would not help her case in the long run. For there would then be another reason to fault her arguments. She would have made her case depend on an interpretation of Parfit’s story that we are by no means obliged to accept.

On Korsgaard’s interpretation, the youthful husband “expects to change his mind without a reason.”22 But the story does not require this. Indeed, there are three alternative, more plausible scenarios consistent with the story as told.

First, the youthful husband may expect his future self to lose his socialist ideals, so that he no longer has reason to give away his estates. And that in itself enables him to credit his future self with a reason to change his mind. For the lack of a reason to act in one way can give one a reason to act in another way.

Second, the youthful husband may credit his future self with additional reasons to change his mind. He may consider the likelihood of his future self’s having a large family to support, or of the peasants having moved away. These possibilities also enable him to credit his future self with reasons to change his mind. Here, it need not be that he has lost his socialist ideals, it is just that he will have acquired new reasons which trump those given by his ideals.

Third, the youthful husband may expect his future self to acquire an allegiance to more conservative views. This would again enable him to credit his future self with a reason to change his mind. Here, he will have lost his socialist ideals, but he will have acquired new reasons which replace those given by his ideals.

In short, it is a possible—not to mention more plausible—interpretation of the story that the youthful husband expects to change his mind with reason, either the reason that he no longer has the reason on which his earlier self acted, or the reason that he now has different reasons on which to act, some which trump his older reasons, others which replace them.

Korsgaard wants to argue that the youthful husband’s recourse to his device—requesting the promise from his wife—is incompatible with an attempt to accord normative standing to the reasons for which he is now acting. In response, we might say that the youthful husband’s recourse to this device is not simply consistent with the attempt to accord such standing to his reasons for acting at Request Time, but part of what maintains them as reasons for him.

To see why, consider a comment made by Korsgaard. She notes that, by asking for his wife’s promise, the youthful husband fails to “will a law that he thinks he can commit himself to acting again later on, come what may.”23 But this is a strange complaint. The law he wills is something he thinks he can commit himself to acting again later on, at least for some indefinitely long stretch of time. Over this time, he thinks he can stand by the contract he has drawn up, by directly resisting all attempts to persuade him to revoke it, for example. He also thinks that there may well come a time when indirect means will be necessary to prevent its being revoked; that is, a time when he will himself move to revoke it, and it is only his wife’s promise not to consent to this that prevents its being revoked.

We can strengthen this objection. The youthful husband might ask how else he is to accord normative standing to his reasons for acting at Request Time unless he deploys a device of the sort he sets up. He might continue: it is surely required of him by the Kantian view of acting for reasons that his way of dealing with his recidivism is not to hide it from himself, or attempt to disguise it in some way, but to ensure that he takes it into account in whatever laws he wills for himself. The “law” he wills at Request Time is that whoever comes into the estates, those estates be given to the peasants. And the purpose of the device is precisely to ensure that this law will continue to be acted on, “come what may.”

Korsgaard can reply that there is another option for the youthful husband: to will a law for himself that both his present and future self can endorse. But we have already seen the problem with this claim. It is at least consistent with Parfit’s story—and may be the more plausible interpretation of it—to deny that this middle course is an option for the youthful husband. The man may be like Luther at the Diet of Worms. Renouncing his estates matters to him not just so much, but in such a way, that he cannot do anything else but renounce them. And the fact that doing anything else is simply not an option for him might naturally be taken precisely as a mark of his integrity.

If this is plausible, then just as before, the Unity theorist with Kantian views appears to be in an awkward position. The fact that the decision not to renounce his estates is not an option for him is plausibly to be taken as a mark of integrity. Yet, in being unable to take a course that his future self could also endorse, it seems that he cannot give normative standing to that for which he acts (under the Kantian conception of what that requires), and thus cannot be a person of integrity (under the Unity criterion). How could we make this conclusion seem defensible?

As before, one line alone seems available to the Unity theorist. If the youthful husband is indeed to be a person of integrity, he must defend and promote his own unity and integratedness over time. And that in turn means he must choose the course that his future self could also endorse. But we saw that this line is self-defeating. For if the youthful husband were to promote unity over and against what he feels he has no option but to do, he will be setting himself “at war” with himself, and thus lose whatever unity and integratedness he has.

Thus we are stuck again in the awkward position. We have some reason to regard the youthful husband as a person of integrity. But if we retain the Unity criterion, at least in the way Korsgaard defends it, we must deny this. Acting with integrity is impossible for him. Again this is precisely the conclusion Korsgaard herself draws. But it is for reasons that may—perhaps should—make her uncomfortable too: that it is not possible for the youthful husband to remain a person of integrity and care so much and in such a way about something that one course of action and no other is an option for him.

6. Conclusion
I have been discussing Korsgaard’s arguments as part of a more general inquiry into criteria of integrity over time. What should we conclude?

I hope to have shown that this option is, if not untenable, certainly implausible: (i) retain the Unity criterion of integrity, (ii) retain Korsgaard’s Kantian views about marriage and reasons for action; and (iii) deny that someone who cares so much and in such a way about something that one course of action and no other is an option for him could be a person of integrity or act with integrity.

Amongst alternative options for which our arguments do provide grounds is this:

(i) accept that someone who cares so much and in such a way about something that one course of action and no other is an option for him could be a person of integrity and act with integrity; (ii) accept that the youthful husband may be someone who cares so much and in such a way about something that one course of action and no other is an option for him; (iii) renounce the Unity criterion.

The claims (Bi) and (Biii) could be linked: perhaps we are directed to endorse such a person’s claim to integrity precisely because we are being guided by adherence, not to the Unity criterion, but to an alternative conception of what integrity is, of what is essential to it. For reasons discussed above, it is the Authenticity criterion that would be the most natural replacement criterion here. I have tried to show in section 2 that it is the one criterion which could be made straightforwardly to support the possibility which (Bi) and (Bii) hold open: that the youthful husband may be a person of integrity.

There is a third option that could also appeal to our arguments for grounds:

(i) accept that someone who cares so much and in such a way about something that one course of action and no other is an option for him could be a person of integrity or act with integrity; (ii) accept that the youthful husband may be someone who cares so much and in such a way about something that one course of action and no other is an option for him; (iii) reject those aspects of Korsgaard’s interpretation of Parfit’s story which block the possibility of (ii); (iv) retain the Unity criterion of integrity.

Does endorsing claim (Ciii) entail rejecting Korsgaard’s Kantian views of marriage or of reasons for action? I think not, for reasons discussed above. I have tried to show in sections 4-5 that it would be consistent with these views to regard the youthful husband as committed in marriage to his wife and as according normative standing to that for which his future self will act. Hence Kantian views of marriage and of reasons do not block what (Ci) and (Cii) hold open: that the youthful husband may be a person of integrity.

Since (C) is the most modest and irenic of the options supported by our arguments, pending further investigation of alternative criteria of integrity over time, it is this option we have most reason to endorse.24


Lecture presented as the Annual Lecture for the Harvard Philosophy Club

Cambridge, Massachusetts, April 2011

1. Dinner at All Souls

I sat at dinner at All Souls College, Oxford with a young philosopher a month or so ago. He said: “I hear you gave a presentation on torture.” (There had been a discussion at the Philosophy Faculty of my recent book, Torture, Terror and Trade-offs: Philosophy for the White House.)

He said, in good humor, “I take it you are against it.” I said, “Yes.”  He said, “Torture is wrong in all circumstances?” I said, “Yes, that’s what I think.” He paused and then he said, “You can’t be serious.” I said, “I am.” The good humor evaporated. He said, “How can it be wrong in all circumstances? Surely we can figure out some combination of possible circumstances where the evil averted by torturing someone would be so much greater and so much more extensive than the pain suffered by the victim of torture.” I said, “No, I think it’s wrong in every case.” He said, “How can you say that, in advance of some extreme cases that might come up.” “Like what?” I said. And he proceeded to set out the familiar ticking bomb scenario:

A nuclear warhead has been planted in London, and it is set to go off in twelve hours. We have one of the people we know planted it, but he won’t tell us where it is so we can disarm it. We think we can force it out of him, but we will have to torture him to do so, and this will undoubtedly involve great pain—inhuman pain—and indignity for him. But, in the balance, think of the pain and carnage we will avert. If the bomb goes off: hundreds of thousands of people will be killed, millions maimed and irradiated, and the site of one of the greatest cities in the world will be poisoned for generations. Wouldn’t it be right to torture someone to avoid this catastrophe?

I said, “No, I think torture is always wrong.” He repeated the hypothetical, multiplying the number of people killed by the bomb by ten, making half of them New Zealanders, including my son and all my relatives, doing whatever he could to up the stakes, to see if there was a point—a threshold, as we say—where my deontological absolutism would crumble. He had no trouble doing this, because a philosopher can stipulate his hypothetical example however he likes. And I had no trouble resisting it, because all I had to do was give the impression of listening carefully, but still just repeat, every time he paused for breath, “No I think torture is always wrong, even in this case.” Eventually, he looked at me as though I were mad—some sort of irrational fanatic—and turned away to talk to the classical historian on his left at High Table.

2. Ticking bombs and other hypotheticals

The use of these hypothetical examples has been quite common in recent debates about torture, although as you know the issue has not in the last ten years been purely hypothetical. American intelligence officials have tortured people—in some cases they tortured people to death—in order to obtain information that would promote homeland security or the security of our armed forces in Iraq and Afghanistan. Alan Dershowitz used a version of this when he mooted his idea of judicial torture warrants: a chapter in his 2002 book Why Terrorism Works is entitled “Should the Ticking Bomb Terrorist be Tortured?”

The tradition of using them goes back at least to Jeremy Bentham, and I will say something about Bentham’s construction of them a little later in this talk.

But it is not only torture. Elizabeth Anscombe reports that in discussion at the Voltaire Society at Oxford when her 1958 paper on “Modern Moral Philosophy” was read, “this case was produced: a government is required to have an innocent man tried, sentenced and executed under threat of a ‘hydrogen bomb war.’”

(This relates to the paradigm that she was considering in her discussion of moral absolutes: not torture but the procuring of the execution of a person well known to be innocent).

Some of my colleagues in the torture debate have condemned the use of these hypotheticals as corrupting—David Luban’s work is a good example.

They make honorable points. Once one starts using these examples, the justification of torture—indeed, the justification of anything—is a matter of simple arithmetic coupled with the professor’s [Parfit-like] ingenuity in concocting the appropriate fact situation.

Many of these ticking-bomb hypotheticals use unrealistic stipulations about knowledge and consequences and, as Henry Shue has argued, many of them assume a clinical image of torture—utterly unrealistic in the light of what we know about torture’s connection to other forms of violence and depravity and about its tendency to metastasize and escape legal and political control.

But there are some real life cases. In years past, the Israeli security forces have sometimes received reliable information that a suicide bomber has entered Israel proper from the territories and have sometimes believed also that people already in their custody might be induced by coercion to give information about his likely destination, information that will enable them to save ten or twenty lives, if not the millions that are imagined by the philosophers.

Today, I am going to assume that we have to entertain these hypotheticals, if not to bully us into accepting a particular policy line, then certainly to explore the nature of the challenge that is faced by anyone who wants to defend an absolutist position on torture. Because even if nothing of practical importance follows from the ticking bomb case, still it seems to have informed the position of most moral philosophers I know, which is that an action like torture cannot ever be ruled out morally for every case or circumstance. And, by the way, that is the sense of absolute that I am using in this lecture. This is not about moral realism or objectivity. At least one of the positions I will examine is non-cognitivist. The question is about deontological requirements or prohibitions. Objective or not, can they be formulated and defended in a way that enables them to withstand the kind of consequentialist pressure that is put on them in these ticking bomb examples?

3. My work on torture

I published a long article about torture in the Columbia Law Review in 2005, and I gave the impression there that I was an absolutist about the prohibition on torture.

But I acknowledged that I had not done anything to argue for such an absolute. My task was to identify and discuss the damage that might be done to a legal system like ours —the trauma that might be inflicted on the law—by requiring it to accommodate the possibility of torture. It was, as I said at the end of the article, second-tier work, compared to the work other people were doing on the wrongness of torture:

The most important issue about torture remains the moral issue of the deliberate infliction of pain, the suffering that results, the insult to dignity, and the demoralization and depravity that is almost always associated with this enterprise whether it is legalized or not. The issue of the relation between the prohibition on torture and the rest of the law, the issue of archetypes, is a second-tier issue.

Charles Fried and Gregory Fried noticed this in their fine book, Because it is Wrong. They noticed that I came close to defending the absolutist position, and that I hinted at it in the language I used—language that was “merely evocative in [my] use of words to describe torture: brutality, breaking the will, mutilating the will, and dignity of the victim,” but that I had “not moved beyond rhetoric to demonstration.”

They acknowledged that it was avowedly not my aim to do so—I was really concerned with this issue about the impact on the law—but they said that it still left a yawning gap in the case against torture.

I did say a little more in a subsequent article in Theology Today on whether the clergy might reasonably be expected to help us with our thinking on torture and complaining about the silence of most of them, during the dark years of 2002-2008; that is also reprinted in Torture, Terror and Trade-offs.

But again that was more flailing around for help than actually explaining how an absolute prohibition on torture might be defended.

What I was finding in my own thought on this matter is that I didn’t have much of an idea of what a moral absolute was supposed to be like and how it was supposed to work. Beyond repeating “Thou shalt not…” several times in very stern tones, and using the evocative rhetoric which the Frieds noticed me using the last time around, how was the absolutist position supposed to be shaped? And what exactly was the response going to be to the accusation implicit in the ticking bomb cases that the absolutist is culpably ignoring or refusing to face up to a whole array of moral considerations on the other side? Does the absolutist just turn away from those considerations (which after all are not just abstract, but human considerations too: the men, women, and children who will be killed and maimed by the bomb if it is allowed to go off)? Does he turn away from all that in the interest, as many people say (admiringly or accusingly), of his own moral purity? Or does he have a way of dealing with the complexities of the situation as the ticking bomb hypothetical presents it? What we want is some sense of moral infrastructure here. And that is what I would like to explore in this lecture.

4. Not just torture

So it is going to be a general inquiry into moral absolutes, though the torture case will never be far from our attention. Anscombe mentioned the example of procuring the execution of an innocent in order to save the lives of many others; so there is another sort of case. She thought that the unthinkability of absolutes in contemporary moral philosophy indicated a startling contrast between the positions taken by most of her contemporaries and what she called “the Hebrew-Christian ethic.”

For it has been characteristic of that ethic to teach that there are certain things forbidden whatever consequences threaten, such as: choosing to kill the innocent for any purpose, however good; … treachery (by which I mean obtaining a man’s confidence in a grave matter by promises of trustworthy friendship and then betraying him to his enemies); idolatry; sodomy; adultery; …. The prohibition of certain things simply in virtue of their description as such-and-such identifiable kinds of action, regardless of any further consequences, is certainly not the whole of the Hebrew-Christian ethic; but it is a noteworthy feature of it; and if every academic philosopher since Sidgwick has written in such a way as to exclude this ethic, it would argue a certain provinciality of mind not to see this incompatibility as the most important fact about these philosophers, and the differences between them as somewhat trifling by comparison

Some of her examples will not be taken up here. I will not explore the issue of absolutes in strictly personal ethics, an issue that she seems willing to explore in her reference to adultery, for example.

Under pressure from the Frieds, I have repented of my sin of concentrating too much on law and not enough on the moral issue. But I am still going to continue to focus on cases which are in some sense about law and policy, even if the question now is what ought to the law to be, or what are the moral considerations that underpin particular legal positions.

My title asked, “What are moral absolutes like?” I shall stick rather closely to that. (But if you are skeptical about moral absolutes, if you are a consequentialist on these matters, you might want to rephrase it as “What would moral absolutes be like, if there were any?”) Even that is slightly ambiguous. I could be asking, “What do (or what would) moral absolutes look like, e.g. propositionally? Are they like rules or are they like other kinds of normative propositions? What normative or evaluative predicates do they use? How exactly is the element of the ‘absolute’ conveyed or formulated?” Or I could be asking that question I intimated before, “How exactly will they operate? What kind of thing does their operation resemble, particularly in their response to considerations on the other side? What actually happens with the facts arrayed against them in the ticking bomb hypotheticals?” I am going to explore both of these versions of my question. Let us begin with the question of what they look like.

5. Analogies with legal prohibitions

We know what legal absolutes look like; and it does no harm to begin with the law, even though we do not want to finish there. In the United States we have a statute forbidding torture. The statute makes it an offense punishable by up to twenty years’ imprisonment (adding that this is punishable by life imprisonment or death if the victim of torture dies as a result).

Now this works as an absolute in our law, simply by virtue of the fact that it is a categorical prohibition; no specific defenses are provided which would reflect the supposed moral power of other considerations, for example in a ticking bomb case. Some have speculated about an implicit “necessity” defense: just as you can sometimes exculpate yourself from a speeding ticket by pleading that it was necessary to travel at a certain speed in order to get an injured person to the hospital in time, so the suggestion has been that intelligence officials might exculpate themselves from a torture rap by pleading necessity to protect the lives of others. But legally it is a very shaky proposition. For one thing, in the international law provisions on which our Anti-Torture Statute is modeled there is an implicit rejection of any exception for emergency circumstances. Many of the rights set out in the International Covenant on Civil and Political Rights (ICCPR), for example, are made specifically vulnerable to derogation: they may be set aside, in times of crisis or emergency threatening the life of the nation.

But that in turn is qualified by a non-derogation clause, which says that no derogation from certain specified articles may be made under this provision, and the list includes the Article 7 prohibition on torture. Effectively, then, the right to torture is made absolute by being non-derogable in contrast to other rights form which derogation is permitted. But in fact, with the Anti-Torture Statute, we don’t need any of this laborious apparatus. As in the case of other criminal prohibitions—on murder, for example, or rape—there is just a straightforward prohibition, which applies a fortiori in all circumstances. There does not need to be anything there saying “in no circumstances” or “whatever the consequences.” The law is simply: “No one is to torture.”

Presumably a moral proposition might be stated simply in exactly the same way. That is, the simplest moral structure is just a set of unqualified “Thou shalt not’s,” standing alone without any addition. If we assume that these do not conflict with one another—for example, because they are all prohibitions on action, a million of which can be satisfied at once—then the obvious inference from their mere presence in a moral code would be that they are to operate categorically and without exception. That would go without saying. And maybe in a moral code these are just basic, as prohibitions are the basic building blocks of law. Maybe we just begin our morality with simple categorical rules like the rule against torture.

You will respond: “Yes, but the prohibition cannot be the starting point. There must be a reason why torture is prohibited.” But reasons come to us already bearing normative shape. Reasons are not just facts; they are facts made relevant by something like a want, an interest, or a principle. Some reasons come to us in the form of value considerations: this kind of feature—suffering for example, counts against an action. But must we assume that all reasons have the same shape and that they always enter into our morality in a consequentialist way? Maybe some of the basic reasons are deontological, as others are teleological. I don’t think we can reject that possibility out of hand, not without begging the question.

A couple of other points about the legal analogy—and remember that at this stage we are just asking: what do moral absolutes look like? How do they present themselves propositionally? Are they going to have anything like the same look, feel, or structure as legal absolutes?

I actually misspoke when I said that our anti-torture statute contains a simple categorical prohibition. It does not. The Anti-Torture Statute does not say, “Thou shalt not torture.” Like almost all criminal law norms it just defines an offense and tells you what the punishment is. You are supposed to infer from that that this something that is not to be done. (Legal philosophers vary in their analytic estimation of the significance of this form. Hans Kelsen thought it was tremendously important. H.L.A. Hart thought it was not. Fortunately we do not have to decide that in this discussion.)

For the moral context, this emphasis on penalty brings to mind the difference between (i) a moral argument for not doing X and (ii) a moral argument for preventing or punishing the doing of X. Plainly (i) and (ii) are not the same. When we say there is an absolute prohibition on torture, are we committed to recommending morally that torturers should always be punished in all circumstances? Maybe not. We do not have to buy into the position of (say) John Stuart Mill, who said, in one of the later chapters of his Utilitarianism, that “[w]e do not call anything wrong, unless we mean to imply that a person ought to be punished in some way or other for doing it.”

And anyway even Mill thought that this definitional feature of duty might be satisfied by the mere compulsions of conscience.

I mention this only because a number of people have toyed with the idea that torture might be absolutely prohibited and yet, if we find someone who has conducted torture in response to the pressures of a ticking bomb situation, we might reasonably refrain from denouncing him or urging his prosecution. We might leave the wrongness in place as matter for him and his conscience—he might have revealed himself to be one of those who, in the words that Weber attributed to Machiavelli, preferred the salvation of his city to the salvation of his soul. 

I am not saying I embrace that position; but it is important to be clear what the operational effect of an absolute prohibition is supposed to be, morally speaking.

6. The formulation of an absolute prohibition

Let us turn now to the words in which an absolute is formulated. When John Finnis writes about moral absolutes, he refers to “norms which have a precise content which is immutable and unconditional.”

A precise content? Does this mean that a moral absolute has to be simple and uncomplicated—three or four one-syllable words like “Thou shalt not kill” in the second table of the Decalogue? I see no reason why we should agree to this.

The U.S. Anti-Torture Statute reminds us that, in law at least, absolutes can be quite complicated. For one thing, torture has an elaborate definition. To be condemned by the legislation, the action must be (a) by a person acting under the color of law, (b) upon another person within his custody or physical control, (c) be specifically intended to inflict severe physical or mental pain or suffering, and (d) not be intended to inflict pain or suffering incidental to lawful sanctions.

The definition in the U.N. Convention against Torture, which is also binding law, adds a fifth element: (e) the act must have a specific motivation, such as obtaining from him or a third person information or a confession.

That whole complicated thing may be what is intended as an absolute. Any act that satisfies these five conditions is absolutely forbidden.

I don’t think it is a cheat to build all this into our analysis. Moral absolutes may be precise in Finnis’s sense, but there is no reason why they have to be simple. It would be a cheat if we built into the act description every single thing about the circumstances of a forbidden act-token—the conditions of its performance, and its short-, medium- and long-term consequences. Then, I think, everyone would count as an absolutist. But the statutory formulation avoids triviality, by identifying some conditions rather than others, and some consequences rather than others, and some motivations rather than others. It does so in complicated but finite terms, specified in advance. When those conditions and consequences are present, then the suggestion is that the act is forbidden, whatever other conditions may surround its performances, whatever other intentions or purposes are associated with it, and whatever other consequences may accrue.

This is true in law and I think it can be true in morality also. It is no part of the absolutist position that absolutes have to be simple. But provided we are still working at the general level of act-types, then a prohibition can meaningfully be called absolute—or its absoluteness can meaningfully be debated—despite the fact that it is quite specific in its characterization. Opponents of the absolutist position should not think they can win the debate simply by challenging our ability to come up with a single predicate or a simple set of Decalogue-style predicates associated invariably with wrongness.

Moreover—and still sticking with Finnis’s emphasis on precision—it seems to me that although sometimes the terms used in an absolute may be simple terms, still it is possible that they are vague and difficult to define, or they may be defined using vague terms (like the use of the phrase “severe pain” to help define “torture” in the American statutory definition). My philosophical interlocutor at All Souls thought this was conclusive against absolutes. (We did actually continue the conversation). He used a Derek Parfit sort of example to question whether I could define an exact threshold of pain where the prohibition on torture would kick in. (Parfit imagined a case in which each of a thousand torturers presses a button and each button causes an imperceptible increase in the victim’s pain; the result is that he ends up suffering excruciating pain. Is it possible to say whose button caused the victim’s maltreatment to pass from something short of torture to torture?).

This, my interlocutor thought, showed that the prohibition on torture couldn’t be a moral absolute. I said I didn’t think it showed that at all. It showed that the prohibition might be difficult to apply in borderline case, but still in clear cases the question of whether something was torture or not (and thus absolutely prohibited or not) could be clearly stated and answered. I think my All Souls friend believed also that the postulated existence of a bright-line threshold on a continuum of severity would tend to discredit the idea of an absolute, by suggesting that something so morally striking could just come to apply owing to a very slight shift along the continuum. The moral relevance of a continuum of severe pain, inflicted by a tormentor, would surely be better captured by a model of balancing the pain caused at any point on the continuum against the pain that we hoped to avert by tormenting our victim. It is a fair point, but it does not quite address the situation where we acknowledge a gray area where there is uncertainty whether a moral absolute applies and yet point to a clear and central case where (we say) it undoubtedly does. The gray area may be quite large. (And, by the way, the gray area is often covered by other morally stringent requirements such as the requirement against hitting people, or more strikingly the requirements forbidding cruel, inhuman, and degrading treatment, whether that treatment amounts to torture or not).

A related question is about the use of morally loaded terms in the formulation of the prohibition. Consider one of Anscombe’s examples: deliberately setting out to procure the execution of someone widely known to be innocent. This is one of her “Thou shalt not” absolutes, but “innocent”—one of the terms in which it is formulated—is hardly a morally neutral term. The description of the action supposed to be prohibited is already morally loaded. Are we not in danger of tautology here? Is “Don’t execute the innocent” not equivalent to “Don’t execute those who are not supposed to be executed,” just as Anscombe’s other example, “Don’t commit adultery” can be read tautologically as “Don’t sleep with those you are not supposed to sleep with”? Well, not quite. “Innocent” has a specific meaning relating to the commission of an offense. So: “Don’t procure the execution of the innocent” means something like “Don’t procure the execution of those who have not committed an offense.” It may be morally loaded; but its moral loading does not prejudge the force of the overall prohibition.

That was a point about thick moral terms in the protasis of a conditional statement—“If someone is innocent…” What about the apodosis, the conclusion of the conditional? Is that always a straightforward prescription—“…then don’t execute him”?  When we talk about moral absolutes, are we always talking about norms framed using the most direct normative language—“Thou shalt not…” or “This ought not to be done” or “This is wrong”? Consider this passage, again from Anscombe:

[I]f a procedure is one of judicially punishing a man for what he is clearly understood not to have done, there can be absolutely no argument about the description of this as unjust. No circumstances, and no expected consequences, which do not modify the description of the procedure as one of judicially punishing a man for what he is known not to have done can modify the description of it as unjust. … [T]his is a paradigm case of injustice.

Is this the assertion of a moral absolute—“Judicially punishing a man known to be innocent is unjust”? Or is that too easy for the absolutist? Do we perhaps not even get to a candidate for moral absolute until we add: “And because it is unjust, it ought not to be done, whatever the circumstances”? I am not sure how to answer this question. It depends, I think, on how the challenge to the idea of moral absolutes is being posed.

7. Rights and violations

A similar point can be made about rights. We say not only that torture is wrong but that people have a right not to be tortured: it is outlawed, for example, in Article 7 of the ICCPR. Now suppose the opponents of moral absolutes—the ticking bomb crowd—believe that sometimes torture is permissible, despite what Article 7 says. They believe that it may be morally right to torture, even though people have a legal right not to be tortured. Might they also say this about moral rights? How would the consequentialists—the opponents of moral absolutes—want to phrase their position? Do they want to assert a moral equivalent of Article 7 and then say that sometimes torture is justified even though it infringes individual rights? In other words, do they want to say that moral rights and moral rightness might come apart? Or do they want to deny that, in those circumstances, people have a moral right not to be tortured? I do not particularly care either way. I am just reminding them of how complex the moral landscape is, and putting them to their choice.

Or think of another idea that might be deployed—the idea of a violation. We might say, “To torture someone is always a violation of him and his moral personality.” Does the opponent of absolutism have to deny that? Can he not just say, “Yes, yes, it’s a violation of him and his moral personality. But sometimes because of this ticking bomb situation, that is what has to be done. Sometimes we have to grit our teeth and violate people.” Is that a coherent position?

There has not been much discussion of the idea of violation.

There is an interesting discussion by Arthur Applbaum on this matter.

Applbaum makes the point that to violate a person is not the same as violating a norm regarding a person (though both are no doubt important); but you can violate a norm regarding a person—for example, by lying to them—without violating them. On the other hand, for present purposes we may want to focus on the idea of violating a person. Applbaum gives the example of rape: “to rape a woman is to violate her, quite apart from violating her right not to be violated.” The person—the victim of rape—is violated; this is something that happens to her.

Her body is violated. She is torn apart by the rape, she suffers the intrusion of another’s will and body, she is used in the most intimate and devastating way. The phenomenology of rape has been developed in feminist writing and I refer you to that.

I want to be very careful in characterizing and developing this point, because this is not an easy subject to deal with properly, and for many people it is not an easy subject to hear discussed in the context of a glib philosopher’s hypothetical.

Bishop Butler reminded us that everything is what it is and not another thing. Rape is rape; and torture is torture. Though it is worth adding that rape is often torture and torture often involves rape, or at any rate involves pornographic and prurient aspects of defilement. (This is an aspect of the corruption that torture sometimes involves; and it is also an aspect of the violation that torture sometimes involves). 

Still, rape is rape and torture is torture, and little would be gained by saying that the two are moral equivalents.

Nevertheless, light may be shed on the wrongness of torture and on the character of the norms that define it as wrong, by careful (albeit partial) analogy with rape. For torture surely involves a similar sort of assault, not just the violation of a legal rule, but the comprehensive and intrusive violation of the body and person of another. And it would be interesting to consider whether the opponent of absolutism has to think of himself as denying this. Can he say: “Yes, I acknowledge that in all circumstances, torture is the comprehensive violation of a person. It is simply that in some cases such comprehensive violation is necessary”? Or does he intend the ticking bomb hypothetical to drive us away even from the perception that torture is the violation of a person? And the absolutist, for his part: can he rest content with the acknowledgement, that in all circumstances, whatever the consequences, torture is a moral violation of a person? Or does he necessarily have to drive towards the hard prescriptive conclusion: “… and therefore it should never be allowed to happen”?

Let me take this one step further. In debates about torture, I have often found myself saying—by way of appeal to my interlocutors, the consequentialists, the ticking bomb crowd—that surely some things are absolutely ruled out as wrong and giving rape as an example—or if my hard-ass opponents shrug that off, the rape of a child (to persuade the child’s mother to tell us where the ticking bomb is). It doesn’t change anything. Of course you know how these things work. We talk as though the methodology used here is that of intuition. But really there is not an appeal to anything properly described as “moral intuition” at all. Don’t be fooled by what your professors say: in modern moral philosophy, as a matter of professional practice, intuition is just the disposition to say something in order to sustain a view. So I bring up the rape example. And the consequentialist, if he wants to remain a consequentialist, just shrugs it off. He doesn’t consult anything, like conscience or intuition (whatever that is). He just says, “Sure, if the stakes are high enough, nothing is ruled out—torture, rape, rape of a child, whatever….” But I watch his face. He is the tiniest bit shaken, and he starts to look shifty, he feels sure such extreme measures would not normally be necessary; he hopes his mother is not listening. This is a case I think where the facial description of the action in question as the deep and comprehensive violation of a person does surely give us pause.

8. How absolutes operate: simply ignoring the consequences?

I said that a second question was this: how are these absolutes supposed to operate (or how would they operate if there were any)? Here I have in mind particularly, the question of how a proposition forbidding an action or characterizing it in negative terms as an injustice or a violation deals with the considerations that are arrayed on the other side in the ticking bomb case.

I suppose one possibility—a sort of limit for our inquiry—is that, having set one’s face against torture or whatever the forbidden action is, one simply ignores the considerations on the other side, refuses to deal with them at all. R.M. Hare, a former holder of the White’s Professorship of Moral Philosophy at Oxford, imagined a figure whom he called “the fanatic.”

The fanatic was someone willing to commit himself prescriptively to certain actions irrespective of their net impact on serving human wants or interests: he just has this ideal, his committed to it, and in light of that commitment he is willing a priori to accept all and any consequences of its universalization. The cases that Hare imagined included a fanatical racist, who would maintain his racism even if his friends, relatives, loved ones, or even he himself turned out to belong to the disfavored group. He is unmoved by these considerations of the universalization of his ideal. In Hare’s words, the fanatic “does not mind if people’s interests—even his own—are harmed in the pursuit of” his ideal.

Hare, who himself favored a sort of utilitarianism so far as substantial moral commitments were concerned, conceded that there might be nothing to say that could dislodge a fanatic from his position. He was inclined to doubt whether fanatical pursuit of an ideal really counted as moral commitment—there was a debate that used to go on in the 1960s about the exact import of the meaning of the word “moral” (moral principle, as opposed to some other sort of principle)—but at any rate Hare conceded that there was no formal irrationality in the fanatic’s position. He could not be shaken from it by argument. No matter what fancy hypotheticals we dreamed up, the fanatic could always respond by saying hard-hearted things like, “Then they too should be denied a vote if they turned out to be black,” or “Then they too should go to the gas chambers if they turn out to be Jewish.” Hare thought we should simply disengage from such people, warn others against them, and comfort ourselves, in the midst of our shared utilitarianism, that we were not fanatical like they were.

Usually we don’t think of the opponents of torture as fanatics, and Hare himself did not make the connection. He did not cite as an example of fanaticism someone who would oppose torture in all circumstances irrespective of the human cost (i.e., irrespective of the interests of the people who were going to get blown up by the ticking bomb). He did once apply the designation to someone who believed in the absolute sanctity of contracts; that was in the context of a discussion of the parable of the unforgiving debtor in Matthew 18: 23-34. Such a person, he said, might be fanatical in his allegiance to “an ideal of abstract justice, of the fiat justitia, ruat caelum variety.”

But the connection here was made, once again by Elizabeth Anscombe, in “Modern Moral Philosophy.” As she deplored the rejection of the possibility of moral absolutes by most of her contemporaries, she had this to say about Hare’s position:

Mr. Hare, for example, while teaching a philosophy which would encourage a person to judge that killing the innocent would be what he “ought” to choose for over-riding purposes, would also teach, I think, that if a man chooses to make avoiding killing the innocent for any purpose his “supreme practical principle,” he cannot be impugned for error: that just is his “principle.”

And I think that is the position. The opponent of absolutism has no choice but to concede that the absolutist position is coherent. The only logical constraint on the possible shape of a moral prescription is that it can be followed (“ought” implies “can”)—so any package of actions and omissions that can be followed can form the subject of a valid prescription. Most consequentialists should not try to deny that one can be an absolutist in this sense: it does not involve one in any contradictions. It just seems hard-hearted when you consider all those innocent people scared and crying, strapped to the ticking bomb.

Now Hare is a well-known non-cognitivist.

He believed that moral propositions do not purport to be descriptive of any reality; they are like universalized imperatives rather than indicative statements. He himself is a utilitarian, but his recognition of the possibility of deontological fanaticism is part of the price he pays for a meta-ethic of prescriptivist voluntarism. Ultimately we all choose our moral positions, according to Hare. I guess most people might be unwilling to simply posit this fanaticism as their own voluntarily chosen position. I don’t go around saying at High Table at All Souls or anywhere else that I just happen to have chosen this absolutism about torture. But then again, that is not necessarily something a consistent non-cognitivist has to be willing to say.

9. God’s fanaticism

Some people adopt the absolutist position by attributing the fanaticism to God: they are simply echoing divine commands which tell us that torture is out of the question, just like idolatry or adultery or any of the other actions that Anscombe mentioned. The theory would be that God gives these as categorical, unqualified, and unyielding commandments and that is that. The idea of a moral absolute is just one instance of the fact that any set of actions that may be performed can be commanded—and there is no telling what God requires of us. I don’t mean that all God’s commandments have to be conceived in this light. Fidelity to God’s word could have made us consequentialists or threshold deontologists or any kind of meta-ethicist. The New Testament is replete with stories of Christ preaching the infringement of what were previously regarded as categorical and inviolable laws—for example, by healing people on the Sabbath.

But there might be some uncompromising divinely commanded absolutes still in play

Divine command theory is hardly flavor of the month these days in moral philosophy. But even if you grant the underlying theism, this is an unsatisfactory position. We can imagine God commanding all sorts of things. But the thing is: can we make sense of His commands? It is no use saying that it is not for us to do that (“Where wast thou when I laid the foundations of the earth?” And so on.) 

The fact is that God often does provide us with reasons—for example, in Genesis, for the law against killing, the reason is that man was made in God’s image.

And we want to know what that provision would amount to if, not unreasonably, we were to expect reasons here as well. We don’t want to be obeying the commands of a whimsical god.

And the point is that the reasons that we need here need to be reasons that address the sources of normative pressure that bear on our principle against killing or torturing or whatever it is. Remember that our question is: how does a moral absolute work to dispel, disarm, or defuse the sources of that pressure? It is not enough to say, “It just does” or “God does it, we know not how.”

In some cases, the considerations that tempt us away from an absolute might be characterized as unholy: unbridled lust, for example, tempting us away from the rule against adultery. But that is not the case with the rule against torture in the ticking-bomb hypothetical. No doubt some torturers, including those who act in our name, are motivated by unholy impulses of sadism and power. But in the formal presentation of the ticking bomb, the considerations that tempt as away from the absolute are in fact also consecrated by religious norms. They include the lives and wellbeing of many innocent people. So the demand for reasons here is not necessarily a challenge to God from a secular position but a challenge to God’s commandment in this one instance, from the perspective of other values or considerations that God surely also regards as important.

10. The real problem

And here we get to the heart of the problem. The real difficulty with moral absolutes—I guess I should have said this earlier—is not the difficulty of imagining the commandment. It is the difficulty of making sense of our sticking with the commandment even when so much of value is stacked on the other side. That is the importance of the products of the consequentialists’ fevered imagination. Their hypotheticals, ticking bombs and all, give us a vivid sense of an overwhelming array of moral considerations—the lives, the suffering of all those who will be affected by the bomb—that need somehow to be displaced from the forefront of our responsibility if a moral absolute is to function. So when I ask, how are moral absolutes supposed to operate, answering, “Just like a command” is not enough. We need to know how does the alleged moral absolute dispose of or relieve the burden of the considerations that the consequentialists seem, quite reasonably, to find overwhelming.

One way it cannot do that is simply by intensifying the stakes on the side of the prohibition. It is no good offering rich and intense description after rich and intense description of what is awful about torture—whether it is the notion of the brutality and the mutilation of human agency that I talked about in my 2005 article, or the idea of the violation of a person that I was toying with in section 7, or the notion that torture offends the bedrock premise “that every human being is a locus of inestimable value” and that it does so by distorting, destroying, and impairing the physical envelope that contains, enables, and expresses the person’s soul—this is language from the Frieds’ book, Because it is Wrong.

I do not mean that it is not important to say things like that—it is—but they do not address the central difficulty of explaining how an absolute works.

This is because the opponent of absolutism, if he knows his business, can duplicate anything we want to say about the badness of torture on his side of the argument. We have Jeremy Bentham to thank for this difficulty. Bentham was an ingenious contriver of ticking bomb hypotheticals. Consider this one:

Suppose an occasion to arise, in which a suspicion is entertained … that at this very time a considerable number of individuals are actually suffering, by illegal violence inflictions equal in intensity to those which if inflicted by the hand of justice, would universally be spoken of under the name of torture. For the purpose of rescuing from torture these hundred innocents, should any scruple be made of applying equal or superior torture, to extract the requisite information from the mouth of one criminal, who having it in his power to make known the place where at this time the enormity was practicing or about to be practiced, should refuse to do so.

Here it is not just lives at stake, as in the common-or-garden ticking bomb example. What Bentham does is wait until the absolutist has given his best and most horrifying characterization of the evil that he has set himself absolutely against, and then Bentham takes that very description and puts it in the scales on the other side. Torture is forbidden because it is so brutal; but what about instances where it is necessary in order to prevent a hundred cases of exactly such awful brutality? Violation is awful, but what about cases where violation is necessary in order to prevent a large number of exactly such violations? Distorting, destroying and impairing the physical envelope that contains, enables, and expresses a person’s soul is awful, as the Frieds have explained; but what if that is the only way we can prevent some gang of Benthamite criminals from distorting, destroying and impairing the physical envelopes that contain, enable, and express the souls of hundreds of people.

I don’t pretend it is easy to come up with the exact screenplays for these hypotheticals. But the broad formula is straightforward. You take the absolutist’s favorite description of the thing he thinks is absolutely forbidden and you postulate the performance of that by someone else to a large number of people, adding the stipulation that it can be prevented by doing the forbidden act in just one case. It is just algebra. All the rest is cinematic detail.

To meet this challenge, then, we cannot just keep coming up with better and better (or rather more and more horrific) characterizations of the allegedly forbidden act. Instead we need to explain how those who take seriously the alleged prohibition are to be relieved of the burden of responding to the cases on the other side, cases that seem to involve exactly the same concerns as those that motivate the absolutism. If we cannot do this—we the absolutists—then we are guilty of not just fanaticism but fanaticism that cannot even keep faith consistently with the values we are supposed to be fanatical about. You see I want the absolutist to be responding to the hardest challenge that can be posed to his position. That is why I seem to be bending over backwards to make the consequentialist critique the best it can be.

11. Possible Accounts

For the remainder of this lecture, I want to consider a number of different ways of responding to this challenge, and relieving the absolutist of this burden.

(i) “You leave it to God”

The first possibility gives God one last chance. I do not know whether you have seen the Clint Eastwood movie, Million Dollar Baby.

There is a character played by Hilary Swank, a woman boxer who is so badly injured in the ring that she is paralyzed and wants to die. She asks Frankie Dunn, her trainer and father figure (played by Mr. Eastwood), to help her die, and he goes to a priest to ask for advice. The priest, predictably, tells him he cannot do this:

Frankie, I’ve seen you at mass almost every day for twenty-three years. Only person who goes to church that much is the kind who can’t forgive himself for something. Whatever sins you’re carrying, they’re nothing compared to this. Forget about God or Heaven and Hell. If you do this thing, you’ll be lost; somewhere so deep you’ll never find yourself again.”

Frankie says: “All she wants to do is die, and all I want to do is keep her with me. And God forgive me but it feels like I’m committing a sin by doing it. By keeping her alive, I’m killing her. How do you find your way out of that?” And the priest says: “You don’t. You step aside, Frankie, you leave it with God.”

That is one way of thinking about how the moral absolutist position takes care of the considerations on the other side. Leave it to the issuer of the command—to the law, if it is a legal prohibition; to God if it is conceived as a divine command. We have been commanded not to do this, whatever the consequences. All right then, let God or whoever the commander is take care of the consequences. It sounds like a cop-out. In the movie, the Clint Eastwood character responds to the priest: “She’s not asking for God’s help. She’s asking for mine.” Indeed it sounds like exactly the dead end we encountered in section 8. But of course “cop-out” is not quite the full story. There is nothing in the absolutist position that counsels or requires us to turn our back on the circumstances which are tempting others away from the absolute. We must still continue to do everything licitly within our power to avoid the bad consequences, to relieve the boxer’s pain and debilitation in Million Dollar Baby if we can, to hunt down the terrorists and the ticking bomb in the customary hypothetical, or to find and close down Bentham’s torture factory, by any and all means that we morally can, to our utmost effort. There is just this one forbidden method that we may not use, and the idea is that we leave the specific consequences of that inhibition to God.

Does anything like this make sense for someone who does not believe there is a benign ruler of the universe who can be entrusted ultimately to make everything come out all right? Well, there is a negative version of it; I mean the position that cautions people against “playing God” with consequences, choosing who lives and who dies, choosing to inflict suffering here rather than allow it there, thinking that we can be so securely in command of the future and of the future consequences that our calculations are trustworthy. For the sake of argument in the classroom, the hypotheticals artificially stipulate what the consequences will be. And in television presentations, like the Fox show “24,” the dramatist’s omniscience conveyed to the viewers also does the work of an artificial stipulation: we know there is a bomb; we know that this man knows where it is; and so on.

But often the wise response is to buck that stipulation, and to insist that all our experience with the law of unintended consequences counsels us to proceed very cautiously with these cases.

This is especially so in situations where we imagine our calculations being at the mercy of the terrorists who have set up the hard choice: the Nazi commander in Sophie’s Choice, the terrorist leader in Bernard Williams’s case of “Jim and the Indians,” the foreign power in Anscombe’s hypothetical that requires us to have an innocent man tried, sentenced and executed under threat of a “hydrogen bomb war.”

But even if the consequences are not being devilishly manipulated, still they may be tangled and unpredictable in ways that defy our calculative morality.

I think this is what is going on in what we might otherwise call the batty reasoning that Immanuel Kant uses in his late essay “On a Supposed Right to Lie from Benevolent Motives.” Kant says you must not lie, not even to the mad axe-man who comes in looking to murder your roommate (who is hiding in a closet) and asks where he is:

[I]f you had lied and said he was not at home, and he has actually gone out (though you are not aware of it), so that the murderer encounters him while going away perpetrates his deed on him, then you can by right be prosecuted as the author of his death. For if you had told the truth to the best of your knowledge, then neighbors might have come and apprehended the murderer while he was searching the house for his enemy and the deed would have been prevented. Thus one who tells a lie, however well disposed he may be, must be responsible for its consequences … however unforeseen they may have been….

This is as good an example of “bucking the hypothetical” as you are likely to find in the philosophical literature. But Kant has a point. The apparently benevolent violation of the principle of truth-telling is based on a sort of hubris—that I can foresee all the consequences, certainly better than those who formulated the principle, and that I have everything in my view and under my control. Such God-like confidence is encouraged by the way we set up our hypotheticals and perhaps it is a good idea to buck them from time to time and concentrate on what we do have control over, not on distant consequences over which our control may be tenuous.

On the other hand, whether we accept Kant’s position “if you have held rigorously to the truth” then, since you have not taken on the burden of consequential calculations, you can’t be held responsible for the unforeseen way in which events develop, is I guess a matter of debate. Maybe everyone has to think as hard as they can about the consequences of their actions and principled commitment is no excuse for refusing this burden.

(ii) Other people’s responsibility

A second line of possibility looks to the fact that, in most of these hypotheticals, responsibility for the consequences that tempt us away from the moral absolute is properly assigned to the terrorists or whoever, who set the situation up. We are not guilty of the tortures taking place in Bentham’s torture factory; in fact we oppose those on the grounds of exactly the principle we are holding as an absolute and we do everything licitly in our power to put a stop to them and bring the perpetrators to justice. But the prime responsibility for those consequences or for the carnage that might be wrought by the ticking bomb is not ours; it is the terrorists’. (This will not always work. With enough Parfit-like ingenuity we can perhaps imagine natural ticking bombs. Thomas Nagel in a famous hypothetical in his 1979 Tanner Lectures imagined a car wreck on a lonely road which caused great suffering that could be relieved only if we used a mild amount of torture on a child to induce an old lady to give us the keys to her car so we could go for help.)

In general, however, we have to be a little bit careful with these assignments and reassignments of responsibility. In every situation, whether others are involved or not, each of us must accept responsibility for the possible world whose actuality our choices define. Suppose that, in a ticking bomb situation, whether the relevant detainee is tortured is up to me and I choose not to torture him and the bomb goes off (knowing or reckoning with a high degree of probability that if he had been tortured we would have been able to disarm the device). Then I have chosen a world W₁ in which the detainee is not tortured and the bomb goes off over a different world W₂ in which the ticking bomb is stopped but only because the person in our custody is tortured. Both those possible worlds have a ticking bomb that was set up wrongfully by someone else; and for that, someone other than me is responsible. But at the crucial fork in the road, it is my free choice between those two subsequent possible worlds, W₁ and W₂, that settles what will actually happen and I have to take responsibility for that.

There is a further point for the consequentialist, which I am afraid I have to draw attention to, which makes this abdication of responsibility even more difficult to sustain. The usual hypothetical invites me to consider whether I would torture the terrorist in order to get information to disarm the ticking bomb. I played along with that in the last two paragraphs. But one response might have been to say, “Well, no. I would not torture the terrorist. I do not know how to do it, and I would not be any good at it.” But that is kind of beside the point. Apparently, there are plenty of people around, able and willing to torture, if only we—the politicians, the lawyers, the legal scholars, and the voters—judge that it is all right to permit them to do it.

This leaves an important opening for the builders of ticking bomb hypotheticals. As the hypothetical is usually set up, there is a familiar asymmetry between doing and allowing: the proposal is that either we torture or we allow the bomb to explode. We did not plant the bomb; the terrorists were responsible for the doing of that. But by failing to torture the terrorist we have in our custody, we in effect allow the bomb to explode by not taking the only means available to prevent it. And sometimes moral absolutists take refuge in this distinction between doing and allowing. By not torturing, they say, they have not done anything awful, though they will acknowledge that they have allowed something awful to happen. But, they will say, it is only the doing of it that was forbidden to them. That is how many absolutists propose to rescue themselves from the central difficulty we outlined in section 9. But if our choice really is what I said it was a moment ago, then this asymmetry between doing and allowing disappears. The alternative is not that I torture the terrorist but that I allow him to be tortured by the hard men on our side who have the know-how and who apparently have no scruples about their integrity. I allow that to happen and the relevant information is extracted from the terrorist along with his fingernails; or I allow the bomb to explode—two alternative things that I allow, and no asymmetry between allowing and doing.

Indeed we can use the Bentham hypothetical to sharpen the choice even further. In Bentham’s example, remember, there is no ticking bomb, there is just a torture factory where a hundred people are being subjected to just the suffering and violation that it is proposed our hard men should inflict on the one guy we have in custody. So it looks as though our choice is simply between allowing one event of torture to happen or allowing a hundred events of torture to happen. Both are cases of allowing.

Again, I am bending over backwards to state the anti-absolutist position as strongly as I can. But what about a response? Is there an equivocation, here, perhaps on the idea of “allow”? I am not sure; I think there may be. On the one alternative, we allow something to happen in the straightforward negative causal sense of failing to do what is necessary to prevent it. On the other side, our allowing has something more to do with permitting or authorizing. We permit, we authorize, we may even order the hard men on our side to go ahead with the torture. That is something we do, and they act in our name pursuant to that authorization.

It may not be like that, of course. They may go ahead even without authorization and they may be deaf to our protestations and forbiddings. Maybe we have to physically intervene to stop them, just as we are going to have to physically intervene to stop the bad guys in the torture factory, if we get the information from their accomplice in our custody. But here we seem to find once again symmetry in regard to allowing. We either permit—that is, refuse to interfere with—the torturer’s activity or we permit—that is, refuse to interfere (in the relevant way)—with the terrorists’ activity. To sustain any sort of asymmetry here, all the weight would have to go on the fact that on the one alternative the torture being allowed was being done in our name, whereas in the other alternative what was being allowed was something we would utterly repudiate and put a stop to if there were a morally licit way of doing so.

(iii) Tainted goods.

Lawyers are familiar with the doctrine of the fruit of the poisoned tree. If we search a person’s house without a warrant or if we coerce someone during interrogation, the fruits of our efforts are tainted by the process we used to get them, and such tainted fruit may not be presented as evidence in a courtroom.

That is a rule we have, and—as several conservative justices on the Supreme Court constantly remind us—it is a rule that other countries do not have and that we could vary if we liked.

Might there also be something analogous in the moral situation? I think it is possible. This is one of the arguments that I said was worth exploring in that Theology Today piece, because often this poisoned fruit or tainted goods approach has a religious flavor.

Defenses of torture depend upon the belief that certain goods—like safety from being blown up—can be attained by problematic means and that whatever else we might say about a particular means being prohibited, we cannot deny the affirmative value of what its violation might secure. If we might secure safety by torture, then our use of this means might be objectionable, but at least no one can deny the value of the safety. At least lives are saved. But perhaps we need not accept this point about unequivocal value. Perhaps we should say that a violation of the absolute rule against torture means that the goods we secure thereby are objectively tainted on account of the immoral methods used to achieve them. The enhancement to our homeland security may be contaminated by the murderous or torturous methods used to bring them about. And a morally sensitive people will be alert to that fact and perhaps unwilling to shoulder the burden of “enjoying” goods that we have secured in this way. Living safe in the knowledge that our “security” was purchased on the back of a waterboard, at the muzzle of a snarling dog, or at the live end of an electrode is a hideous thing.

Let me put this more affirmatively. A lot of our affection for due process and the Rule of Law is bound up with the character of our enjoyment of the goods and the safety that law provides. Having committed ourselves to the Rule of Law, we know this about the character of the goods we enjoy: they have not been secured for us brutally, arbitrarily, or oppressively. But if we take away the Rule of Law and introduce brutality and the infliction of torment into our national security policy, then the result may be that our personal safety or the safety of our streets becomes a reproach to us, a tainted and clammy form of satisfaction that we can enjoy only with our consciences turned off.

Again, it is not easy to secularize this reasoning, which may seem like a sort of magical thinking. I have attempted to deal with it at length in a long analysis of the nature of security in the Torture, Terror and Trade-offs volume.

I cannot go into that here. But let me try one analogy. It is sometimes said that the goods we aim at by their nature place limits on the means that may be used to attain them. For example, if national independence is attained in an insurgency struggle using a terrorist campaign, then the leadership that comes into power if our campaign succeeds will be a leadership of hard men and assassins accustomed to operating in ruthless and clandestine ways. And it may be not be easy for them to change their habits of thought, conspiracy and action or to sanction the empowerment of those accustomed to gentler methods more appropriate to a free polity.

The insurgents may assure us that the end—national independence—justifies the means. But the end is generic—national independence is a very complex multifaceted good—and its complexity suggests that there are many ways in which it might be affected or modified by the character of the means that were used to secure it. It may come to us in a variety of different characters, depending on the means used to attain it. Some of these characters may be more comfortable for us than others. If we value national independence of a certain kind (for example the national independence of a free and constitutional democracy) then it may be important to prefigure that in the means used to attain it. And there may be no other way of ensuring that we get this version of the end than by placing limits on what we are prepared to do or permit to bring it about.

Or consider this example, more closely related to the war on terror. Suppose we can secure homeland security only by operating a frankly discriminatory policy, subjecting Muslims, for example, to much greater intrusions on their liberty than non-Muslims. That method may make us safer, but it will be safety-in-a discriminatory-polity that we now enjoy, with glimpses of Muslims being rounded up or beaten as the rest of us remain safe. This may not be the sort of safety we were looking for. We now have to think of our safety, not as the safety of Americans, but as the safety of white Americans or else we have to accept the costs and dangers of turning a blind eye to this characteristic of the end we have achieved.

Again, none of this is conclusive. It is just an exploration of the conceptual space for making sense of an absolute.

(iv) The rules of the game

A fourth line of thought is more abstract, and it is developed through an analogy between moral absolutes and the constitutive rules of a game. For this I am indebted to an old article by Nicholas Denyer entitled “Chess and Life: The Structure of a Moral Code.”

Denyer invited us to reflect on the rules of chess and the considerations that might tempt us to violate them in the course of a game. Imagine that you are developing this really interesting line of attack, which will be devastatingly effective as well as fascinating to generations of chess aficionados; but someone points out that in the early stages this strategy involves moving your king into check. And they say, “You cannot do that; it’s against the rules.” You point out all the advantages that might accrue if you are allowed to break the rule just this once, and you say to the person who is restraining you, “How can you be so insensitive to the advantages that might accrue?” But the stickler for the rules will be adamant. The point of the analogy is that in a game we must not let the rules be infected by considerations derived from the goals that a person is supposed to secure in playing the game. We insulate the rules from goal-based considerations, and accordingly we are able to treat them as absolutes.

We must not even use Benthamite reasoning. We must not say, for example, that breaking the rule in this case is a good way of minimizing its violation for other cases: “If I leave my king in check this move I’ll subsequently be able to keep it safer from check than it would be if I now remove it from check.” The point is that leaving the king in check is what Denyer calls a “deontic impossibility.”

It cannot be used as a strategy to minimize check time or to discourage others from moving their king into check or anything like that.

Denyer thinks this is an important result. It shows, he says, that the idea of a moral absolute is not incoherent, even for what we would recognize as the Benthamite cases. But does it really help us with relieving the burden of the consequences on the other side? Denyer assigns those consequences to the domain of goal-based play of the game, but he says they cannot be allowed to infect the domain of the rules of the game itself. That domain-distinction seems fine for chess, but maybe it is a little artificial for the game of life. And the consequentialist about morality, who is unconcerned with chess, may challenge the distinction on that basis.

I am actually not sure what to think about this. The structure that Denyer lays out is more or less exactly the moral structure that Robert Nozick associated with “side-constraints” in Chapter Three of Anarchy, State and Utopia.

There Nozick drew a familiar contrast between that approach to rights and what he sometimes called a “utilitarianism” of rights (which seems to be what is going on in the Bentham hypothetical)

In contrast to incorporating rights into the end state to be achieved, one might place them as side constraints upon the actions to be done: don’t violate constraints C. The rights of others determine the constraints upon your actions. … This view differs from one that tries to build the side constraints C into the goal G. The side-constraint view forbids you to violate these moral constraints in the pursuit of your goals; whereas the view whose objective is to minimize the violation of these rights allows you to violate the rights (the constraints) in order to lessen their total violation in the society.

But even Nozick acknowledges that the fact that this “is a consistent position does not, of course, show that it is an acceptable one.”

And his own rationalization of it commits the fallacy we noted earlier: it points to the awfulness of the violation of an individual when his rights are attacked, saying that this is the only life he has; he has a conception of life’s meaning; he must not be used simply as a means to the well-being of others; and so on.

That is all very fine, until you remember that in Bentham’s hypothetical, something like this can also be said about each of the people who are being tortured in the torture factory, the unique individuality of whose view on life will be vindicated if only we do this one bad thing that puts us in a position to rescue them.

The difficulty is that there does not seem to be anything in Nozick’s structure that can really play the role of Denyer’s impenetrable firewall between the rules of the game and the goals of playing it. Indeed, often in games—rugby is a good example—we place the rules under continual revision to reflect the pressure that they have been put under by players trying to play the game the best they can. (You may remember that in the 1970s, Ronald Dworkin suggested that we might even want to subject the question of the proper interpretation of the rules of a game—I think he used the tournament rules of chess as an example—to input from the players’ perspective of what made for better play or a better tournament).

If there is anything to glean from the Nozick/Denyer position, it may be this. The idea of individuals as inviolable bearers of human dignity may be a natural or objective one; but in social life, to be effective, it has to be something of an achievement, a construction, something we build for ourselves. And maybe this moral construction involves something like a firewall between the goals of moral life and the rules that constitute the persons who are engaged in it. We say definitively,

Here are the ways in which people are to be treated. Here are the rules for treating people; abiding by these rules qualifies you as a player in this game of dignity and respect that we are defining here. And we hold these rules sacrosanct, because we know that once we start messing with them, all bets will be off, and individuals will start once again being treated simply as means to each other’s ends.

Something like that might be going on, and that may yield—constructively—a sufficiently robust version of a firewall between the rules of the game and the goals of play within those rules to enable us to speak confidently of moral absolutes. Only now, the game will be presented not just as a limited and artificial recreation, but as a deadly serious and pervasive basis for all social and political interaction.

(v) “Threshold deontology”

There is one last maneuver I want to consider, though I am not sure whether it is a maneuver for the absolutist or not. But it is in play in the philosophical debate about these matters, and something needs to be said about it or some questions asked about it.

It is a conception called “threshold deontology,” which as I understand it acknowledges treating certain moral rules—such as a rule about torture—as a near-absolute but indicating a willingness to abandon it when the consequences piling up on the other side pass a certain threshold.

So, for example, the rule against torture may not be shaken by a ticking bomb hypothetical that involves the lives of ten people imperiled by the terrorists, but it may be shaken by a hundred or a million.

Nozick famously hinted at something along these lines. He said, of the rules or rights we considered a moment ago, “The question of whether these side constraints are absolute, or whether they may be violated in order to avoid catastrophic moral horror, and if the latter, what the resulting structure might look like, is one I hope largely to avoid.”

There is a hint of a similar idea in some of Michael Walzer’s musings on supreme emergency as permitting actions that would otherwise be forbidden by the laws and customs of armed conflict.

And I guess there is an element of ordinary moral common sense in this as well. For example, Joseph Raz’s conception of the exclusionary reason associated with promise-keeping is supposed to mean that although promises may not be broken whenever there seem to be good reasons for doing something other than what has been promised, still the exclusionary apparatus is properly held vulnerable to considerations of overwhelming moral importance. We break our promise to save lives, even though we do not break them because lunch with someone else would be marginally more productive.

For our present purposes, the idea is that threshold deontology indicates a way of coping with the burden of the humanitarian considerations that oppose a moral absolute. The threshold deontologist sticks to his principle up to a point, but he is absolved from the charge of utter heartlessness by being willing to switch sides in the stand-off when the stakes on the other side get big enough—that is, when they pass the threshold that his version of deontology enshrines.

I guess the natural question to ask is, “Well, where do we set the threshold?” Or “How do we set the threshold?” or “Who sets the threshold?” Those are all good questions. But we cannot even begin to address them unless we know something of the moral infrastructure. How is this threshold supposed to work? What is its mechanism? What does it operate upon? What is happening when the threshold—wherever it is—is finally crossed?

I am afraid I have not seen any good discussion of this in the literature. The affirmative accounts that I have read simply state the position without defending it: Michael Moore simply announces that:

It just isn’t true that one should allow a nuclear war rather than killing or torturing an innocent person. It isn’t even true that one should allow the destruction of a sizable city by a terrorist nuclear device rather than kill or torture an innocent person. To prevent such extraordinary harms extreme action seems to me to be justified.

But there is no argument, simply the announcement that this is a position that can be taken. Because it can be named, people just assume that “threshold deontology” is a coherent and viable position and that, in reflective equilibrium, it may capture something of our deontological instincts and also our unwillingness to seem too morally idealistic when the stakes are very high. Philosophically this is an unsatisfactory state of affairs, and it is unclear whether we should regard this as really a version of deontic absolutism, with some sort of “out” in the higher register, or really as a modified consequentialism, which is not really deontological in its tendency at all.

It might represent, for example, a sort of rule- or indirect-utilitarianism based on the untrustworthiness of the consequentialist calculations that act-utilitarians have to engage in. At the margins, the calculations of cost and benefit that tempt people away from moral absolutes are pretty untrustworthy and we might be better off sticking with the moral rule.

But there comes a point when the apparent advantages of deviating from the moral rule are so high that they overwhelm any of the uncertainties, and that is what happens when the threshold is crossed. That is one possibility, though why we would call this threshold deontology, I am not sure. Moore seems to be convinced that it is not a consequentialist position at all, and he eschews this sort of justification.

The other possibility is that threshold deontology does represent a real form of deontological absolutism, but is coupled with a recognition that, in the most extreme cases, people’s will to do what morality requires may crumble and effectively all bets are off. It is a little bit like David Hume’s suggestions in the Enquiry about what we now call (following Rawls) the circumstances of justice:

Suppose a society to fall into such want of all common necessaries, that the utmost frugality and industry cannot preserve the greater number from perishing, and the whole from extreme misery; it will readily, I believe, be admitted, that the strict laws of justice are suspended … and give place to the stronger motives of necessity and self-preservation. Is it any crime, after a shipwreck, to seize whatever means or instrument of safety one can lay hold of, without regard to former limitations of property?

Suddenly the demands of morality, which might formerly have seemed sensible, become startlingly unreasonable, and there is no basis on which we can reasonably expect people to abide by them. The threshold is a psychological one—about whether people can pay attention to morality at all. I do not know whether we should acknowledge such a threshold or, if we do, whether we should regard your common-or-garden ticking bomb hypothetical as illustrating its application. But in the interest of trying to distinguish the issues that are really at stake in the debate about absolutism from those that belong somewhere else in moral philosophy, we might want to say that the existence or non-existence of absolute norms within the domain where morality does or should sway, should be treated as quite a different matter from the further and interesting question of whether there are limits to that domain.

11. Conclusion.

It is time to finish. I am conscious that I have not produced any final answer to the question how the absolutist deals with the burden of the humanitarian considerations that seem to motivate the infringement of his absolute principle. If anything, I have made matters worse for the absolutist by indicating—with the connivance of Jeremy Bentham—just how heavy that burden is. My discussion is frankly inconclusive at best.

In section 10, I suggested one or two lines of promising argument, particularly (i) the caution against “playing God,” (ii) an insistence on the proper allocation of responsibility, (iii) the notion of tainted goods, and (iv) the idea of the inviolability of persons as one of the constitutive rules of the constructive moral “game.” But none of these is more than suggestive, though I think their suggestiveness helps enrich the debate.

The last line of argument we explored—(v) threshold deontology—we found pretty much a dead end. It adds nothing in the way of argument. It simply shows that an intermediate position can be stated without incoherence and perhaps also reconciled with the intuitions of some people (namely those who, like Michael Moore, announce it as their position). But we know both from the example of the law and models of morality—divine command theory and the account of the moral fanatic in the framework of Hare’s universal prescriptivism—that various coherent positions can be stated in this matter. The challenge is to make sense of them, and I am afraid that the threshold deontologists have done nothing to make sense of the alternate ways in which they require a moral agent to respond first—and dismissively—to the burden of one set of humanitarian considerations and second—and submissively—to the burden of a larger set. Simply stating positions and comparing intuitions is not going to get us anywhere in this enterprise. Both sides—and the middle—can play that game all night long.

Since no conclusive answer has been furnished to our question—How are moral absolutes supposed to work?—it is tempting to say pessimistically that the absolutist must admit defeat. This is inferring too much, I think, except in the ordinary sense that the absolutist concedes right here and now that he has not done what he would like to have done. But nothing has been done to show that the question we have posed for the absolutist cannot be answered. And, if I say so myself, a certain amount of progress has been made in indicating, first, what such an answer cannot be like, and secondly what it may be like. Both kinds of progress are important—the first, for indicating threads worthy of further exploration, and the second for redeeming us from wasteful excursions that do not really address the heart of the problem.

Putnam: I am, of course, a mathematician as well as a philosopher. And I’m very happy about that [laughs]. I started working in mathematical logic when I taught in Princeton in the fifties. Georg Kreisel, who was really my mentor in logic, was at the Institute for Advanced Studies. I went to see him and told him that I taught what was called symbolic logic in the philosophy department, and that I’d like to see if I could prove an original theorem. I do mathematics rapidly when somebody explains it to me on a blackboard—I can pick up a proof in an hour by listening to someone explain it and asking questions, but it would take me weeks to read it in a published paper. Mathematics papers are not written to be easily understood. There’s a real divide between the way mathematicians explain proofs to one another and the way they write them up.

So Kreisel explained some concepts and then gave me a problem, which I solved, and then he said, “Now you publish this result.” The paper I wrote (“Decidability and Essential Undecidability”) came out in 1957. It was my first mathematical publication. It wasn’t on Hilbert’s 10^th problem, which I went on to afterwards. In fact, our relationship cooled eventually because Kreisel was dead set against my working on that problem. He thought it was a dead end—that nothing would come of it. But around the same time, a person who is still my very close friend, a mathematician named Martin Davis, also came to the Institute for Advanced Studies. We and our wives became constant companions. There was a summer institute in mathematical logic in 1957 at Cornell. The Davises and the Putnams were there together. Martin had published some papers working towards a solution to Hilbert’s 10^th, and I suggested that we try to extend his work and try to go further.

So that was the beginning. We didn’t have much time that summer, because there was a conference on just about every topic in logic going on, but the next two summers we rented places near one another in Connecticut, north of Hartford, and we worked full time. In fact, as Davis described it, I was the idea man and he was the filter. In 1959, Martin Davis and I proved a theorem which, together with results by Julia Robinson and a later result by Yuri Matiyasevich, added up to a “negative solution” to that problem. As Davis describes our way of working in George Csicsery’s film “Julia Robinson and Hilbert’s 10^th Problem,”

I was a “fountain of ideas”, and Martin’s role was to say, “That’s ridiculous…that’s ridiculous…that’s ridiculous…” until every so often one of my idea wasn’t ridiculous, and then we would both set to work on it.

So those summers I would sometimes work on the problem until four in the morning. I was still young, in my very early thirties with lots of energy. And by 1959, we had almost a proof—not of the unsolvability of Hilbert’s Tenth Problem, but of the unsolvability of what are called “exponential Diophantine equations”. These are polynomial equations in which the exponents can be variables, and the coefficients are all integers and one asks for a solution in integers. x³ + y³= z³ is a Diophantine equation, that was proved unsolvable in integers by Fermat himself, and he claimed he had a proof (though no one knows if he really had one) that xⁿ + yⁿ= zⁿ  is also unsolvable in integers whenever n is bigger than 2. Now that’s not a Diophantine equation, because n is variable, so that’s an exponential Diophantine equation. But Martin and I had a proof that the decision problem for exponential Diophantine equations is unsolvable, if the following hypothesis is true, which we called “PAP” for Primes in Arithmetic Progression: there are sequences of prime numbers as long as you like in arithmetic progression. PAP was, in fact, proved in 2004 by Ben Green and Terence Tao. If PAP had really been necessary to prove the unsolvability of exponential Diophantine equations (and thus, necessary to the proof of the unsolvability of Hilbert’s 10^th Problem), then the solution to Hilbert’s Tenth would only have been completed in 2004, and Green and Tao would have to be considered as completing the solution of Hilbert’s 10th.

However, going back to the summer of 1959, we already knew, via a result of Julia Robinson’s, that the unsolvability of the decision problem for exponential Diophantine equations implies the unsolvability of Hilbert’s 10th problem if there is even one Diophantine equation with the following property: it has infinitely many solutions, but as you change one of the coefficients, one of the parameters, it has only finitely many solutions for each value of that parameter, and the largest of those solutions grows roughly exponentially as  the value of the parameter is increased. So at that point, the solution to Hilbert’s 10th Problem had been reduced to two things: first, the problem of PAP—but when we sent our proof to Julia [Robinson], she showed, in fact with a variant of an argument of mine, that we didn’t need PAP! So we did have a proof of the unsolvability of exponential Diophantine equations. And the three of us published that together in the early 60’s in Annals of Mathematics. Then completing the (negative) solution of Hilbert’s 10th Problem was a matter of somebody coming up with a single example of such a Diophantine equation, and within ten years, Yuri Matiyasevich in Russia did come up with such an example—an example closely related to the Fibonacci numbers, by the way. Martin and I had thought that the Fibonacci numbers might hold a solution, but we were unable to prove it. But Matiyasevich did give an example, so now The Gang of Four had a proof of the negative solution to Hilbert’s 10^th.

After that, still focusing on my mathematical work, I turned to set theory. My work was concerned, among other things, with what is called the “fine structure of Gödel’s constructible sets.”

HRP: I think most people are probably less familiar with that aspect of your work. Would you mind describing what that entailed?

Putnam:  Well, Gödel proposed an axiom that he himself rejected later, on philosophical grounds. He didn’t like it, and few set theorists liked it—although it’s consistent with the other axioms of set theory, so you can’t refute it. It puts a certain kind of structure on the universe of sets: it says they’re all constructible, in a sense I won’t try to define [laughs]. But if all the sets there are constructible in Gödel’s sense, then that would imply the axiom of choice, and it would imply Cantor’s famous continuum hypothesis. It would essentially close all the major open questions there are about the structure of the universe of sets—that’s what makes it so interesting. I started a field called the examination of the fine structure of Gödel’s sub-universe of constructible sets, which is still an active field of research. Mainly, contemporary set theory is looking for alternatives to Gödel’s axiom of constructability, but the constructible sets themselves are one of the major tools.

HRP: Some active research in set theory and logic is philosophically motivated—you mentioned that Gödel gave up this axiom on philosophical grounds. Was there any relationship between your work on constructible sets and your philosophical leanings?

Putnam: Not directly, I don’t think, although I’ve thought a lot about the philosophy of set theory, and I am a realist about set theory.

HRP: The topic of realism brings up another question I’d like to ask. A lot has been made about how frequently you’ve changed your position on various philosophical issues over the years, and I want to get to that, but first I’d like to ask, are there any conclusions or arguments or ideas you’ve held firmly onto your whole career?

Putnam: Well, first of all I’d say there are certain themes that characterize my work, throughout my whole career. I think we could say that I have always been seeking a way of doing philosophy that’s scientifically respectable and scientifically informed—in fact, you know I try to be sophisticated about and draw on a wide range of sciences. So I think philosophy should be scientifically respectable and scientifically informed, but at the same time, I don’t think that philosophy is reducible to the sciences. That sort of reductionism, I think, is wrong. Although there’s common ground between philosophical arguments and actual scientific practice, one danger of being too much in awe of science is that you have to steer clear of identifying science with a number of different ideologically driven definitions of science, which have been common in the past 100 and more years. So I would say, ideally, this scientifically informed way of doing philosophy, which doesn’t believe that philosophy is going to become a science or that philosophy is going to be reducible to various ideological notions of science, is a common theme in all my work.

As far as conclusions are concerned, quite a few. I’d say first of all, I’m only aware of two major issues on which I’ve actually changed my mind. One is computer-machine functionalism, or computer-program functionalism—this I’ve certainly changed my mind on—and the other is what I called internal realism, a position I held for about fourteen years. And quantum logic, if you want to count that. But, for example, I think all but about four or five of the papers in my first collection of papers I still agree with. I mean, obviously I think some arguments were good, some arguments I wish I had given, but basically I think I agree with just about every paper in my first two Cambridge University Press volumes, if not with every argument. But what philosopher who lives as long as I have agrees with every argument he ever gave?

To be more specific, let me describe the papers in which I found my own philosophical voice. I came to Princeton in 1953, very much as a student of Hans Reichenbach—very much a logical empiricist. But Reichenbach I see as a very atypical logical empiricist in that he was a scientific realist, and he believed, I think incorrectly, that positivism, or the verifiability theory of meaning, and realism, were reconcilable. That’s all over the place in his book Experience and Prediction, and he also attacked the idea that theoretical entities in science are constructs, which was a hallmark of Carnap’s view. Reichenbach always held that they are not constructs; they’re inferred entities. He held that we make inductive inferences to the existence of things we can’t see.

HRP: Something that you also took over from Reichenbach?

Putnam: That’s right. And I think I took from Reichenbach the idea that it is important to ask what the implications of the great scientific theories are. He was a friend of Einstein’s and of course he knew relativity very well, and in fact I had my introduction to relativity through Reichenbach. But he asked, what are the implications of these new theories—quantum mechanics, though he was less successful there—but also what are the implications of relativity theory, general and special, for classic philosophical questions about the nature of causality and the nature of time? So I would say it’s unfortunate that he made an alliance with Carnap. I think Carnap talked him out of some good ideas, and also made him fit what he was doing into a framework that was incompatible with it. But that interest I kept from Reichenbach—asking “What does contemporary physics have to tell us about the nature of physical reality?”

Now the first paper in which I found my own voice, the paper in which I ceased being a Reichenbachian, was a paper I wrote in about 1958 (although it was published many years later), “The Analytic and the Synthetic.” The day I met my wife Ruth Anna, which was August 27^th, 1960, I gave a paper called “What Theories Are Not,” and in a way, those two papers were my manifestoes. In those two papers together, you can see an awful lot of my later thought. Even some of the ideas in “The Meaning of ‘Meaning’” can be seen in “The Analytic and the Synthetic.”

HRP: If I had to guess, I would have also put “It Ain’t Necessarily So” in that group of early papers.

Putnam: Oh yes, yes, but most of “It Ain’t Necessarily So” is in “The Analytic and the Synthetic.” Of course, “It Ain’t Necessarily So” was written several years afterwards.

HRP: I would like to turn to the issue of realism now. It’s obviously a major theme in your work, whether a defense or a criticism of it.  Now, you mentioned earlier that one of the two things you’ve given up was internal realism. Can you give us a brief history of what happened with internal realism—why you were motivated by it, and why you gave it up?

Putnam: Starting with a Presidential Address I gave here in Boston to the Eastern Division of the American Philosophical Association in 1976, I slipped back into a form of verificationism. Partially that was the influence of Michael Dummett, who had recently given the William James Lectures here. Well, first, my well-known “model-theoretic argument” occurred to me—that was a dilemma for me. The model-theoretic argument drove me crazy for about four years before I gave into it. I spent about four years trying to refute it, which I think can only be done if you reject some of its assumptions, which I did not do. I didn’t see that the model-theoretic argument itself really depended on questionable assumptions. It was an argument that assumed, as it were, that the external world impinges—is available to us, is a constraint on what we can say—only via observation sentences. From Dummett, I took the idea that there’s no privileged observation vocabulary. But I ultimately identified observations with something going on internally. I think that my internal realism was very closely connected to the problem of skepticism—basically, I slipped back into a Cartesian picture: here I am, locked inside of my brain, I have these inputs, and the only thing that can constrain what I say is these inputs plus what I called “theoretical constraints” like simplicity and elegance and so on.  Once you accept that picture, I think then the question of how can a unique interpretation, a unique correspondence relation between our terms and anything external, be fixed becomes unsolvable. And having tried to refute that for four years the wrong way, instead asking myself whether this wasn’t a hopelessly positivistic picture of the mind, I decided “if you can’t lick ‘em join ‘em” [laughs].

Most of the critics of the model-theoretic argument up to this day don’t really talk about the need to have an alternative picture—except McDowell, of course, who later influenced me very much. But by 1990, I had pretty much decided that picture was wrong.  My late “internal realist” books and papers used the term “internal realism” but mainly they talked about conceptual relativity. They said less and less about verificationism.

There was a conference on my philosophy, immediately after I had given the Gifford Lectures [published as Renewing Philosophy], called the Gifford Conference in Saint Andrews in November of 1990, and in some of my replies, particularly my reply to Simon Blackburn, I said that verificationism was a mistake, and I went on to explain why it was a mistake in things that I published over the next four years. There was an issue of Philosophical Topics on Hilary Putnam in 1992 in which in one of the papers I explained why it was a mistake. Unfortunately, that issue is hard to find—I’ve discovered that in almost all the libraries in the world, that issue has been stolen! Every time I visit a university and say “Well, you should read that issue,” the students say “Oh, well that issue’s been stolen from our library” [laughs].

HRP: That might be flattering.

Putnam: It’s flattering in a way, but I wish that the thieves had just contented themselves with photocopies [laughs]. But then in the Dewey Lectures [collected in The Threefold Cord; Mind, Body and World], I did explain why I thought that internal realism was wrong, and I called my new position “commonsense realism”, but I have discovered that this name is a Rorschach test. It was a mistake to use it, because everybody who sees it comes up with a different understanding of what I meant by “commonsense realism.”  I thought it was pretty clear, but I’m not going to use that term any more. Now I just say that I am a realist in metaphysics.

HRP: In the Dewey Lectures, you give a critique of functionalism and representationalism in general, and I take it that such a criticism is motivated by these issues of realism—

Putnam: No it wasn’t. Jerry Fodor and I—he was my student—have, I think in many ways, converged much more than it might seem, although he draws different conclusions. For example, he’s given up the notion of narrow content now, in his most recent book. He rejected the notion of narrow content for good reasons, although he throws out the baby with the bathwater by deciding that we have to throw out the notion of intentional meaning entirely, and only talk about reference, which is not a conclusion I draw. But I would say that the reason [those critiques] didn’t have to do with issues of realism, is—well, first of all, Jerry’s review of The Threefold Cord, was a strange review in that he described the book as an attack on mental representation. I looked in the index, and there’s only one passing reference to mental representation! I mean, just take a look at the book—when an English speaker thinks, yes, there may well be mental representations that are connected with the words she uses, but the question is are there language-independent, innate mental representations? That’s a very different kettle of fish. That wasn’t the topic of The Threefold Cord, which was perception—which Jerry has very little to say about.

Now I think the real issue is this: let’s go back to the “The Meaning of ‘Meaning.’” In “The Meaning of ‘Meaning,’” I did suggest the notion of narrow content; I wrote “psychological states in the narrow sense.”  I suggested there that narrow content might not be useful for psychology, and in fact I don’t think that it’s useful for psychology, and I try to spell that out in my Royce Lectures [reprinted in The Threefold Cord; Mind, Body and World]. But Jerry, for a long time, thought that psychology is about narrow contents. Now he thinks as I do, that that isn’t really what psychology is about—that you can’t put psychology in that procrustean bed. If a psychologist says, the rat saw the cheese and pressed the bar, she’s talking about seeing cheese and pressing bars, which are not inside the rat’s brain [laughs]. To demand that psychologists must never talk about bars and cheese is rather odd. The issue was about narrow content, and the main point about narrow content is that it abstracts from reference, so that the issue of realism doesn’t really arise with respect to narrow contents.

Now, about my criticism of computational models of narrow content. I think that the problem with the notion of narrow content (this is the argument I spell out in about, I think, six pages in The Threefold Cord) is this: we really possess only a sufficient condition for saying that something’s the narrow content of an utterance, and one which is never realized in the real world. Namely, if somebody is in the same brain state as someone who thinks a thought with the same broad content as “The cat is on the mat,” then that person has a thought with the same narrow content as “The cat is on the mat.” But the trouble is that “A and B utter sentences with the same narrow content when they are in the same brain state” is not a very useful criterion of identity. And if you say “Well, it’s enough that he or she be in a brain state such that somebody else could be in it while thinking a thought which has the same broad content as ‘The cat is on the mat,’” then the notion of narrow content then becomes parasitic on the notion of broad content, that is, on the ordinary notion of meaning, because broad content is just a rational reconstruction of the ordinary notion of meaning. I don’t believe we have an independent notion of narrow content. If you could identify narrow contents with computer programs, then you could say, “Okay, there is a computational criterion for sameness of narrow content.” But both Jerry and I have come to the conclusion that there is no computational criterion for saying what is “meaning” in any sense of meaning: broad content, narrow content, or reference.

HRP: I take it that the arguments for that are in Representation and Reality?

Putnam: And now, Jerry also accepts the idea that mental states, intentional states, and meaning-bearing states, are not only compositionally plastic, but computationally plastic as well.

I don’t want to say that sameness of meaning is reducible to a notion of good interpretation (because I don’t think one is prior to the other), but I think that sameness of meaning and reasonable interpretation are entangled notions. That to formalize the notion of having the same content in any of these senses, you would have to be able to formalize the notion of good interpretation—and while I can’t prove that it is impossible, there is no reason to think that it is possible.

I remember once asking Noam [Chomsky], “Do you think that there is a computational criterion for when the brain is acting up to its competence in scientific reasoning?” There is supposedly a scientific competence like our linguistic competence, and all of that, he thinks, is innate in a different way and in a different place than our language organ. And he said “Yes.” But surely one should be able to prove that if our total rational capacity—inductive logic, and deductive logic and mathematics—is formalizable, then we couldn’t know the program [laughs]. I think it’s easy to sketch a proof that that is indeed the case. But maybe someone will show that there’s something different about interpretation, and that interpretation really is formalizable, but all I want to say is that we don’t even have a sketch of an idea how to formalize it, and there’s no a priori reason to think that it has to be formalizable. The example of the Gödel theorem suggests that things are otherwise. In fact, I have a paper that just came out titled “The Gödel Theorem and Human Nature,” that argues that it’s more than just a suggestion.

HRP: I think that people working in cognitive science and psychology in general find claims like that a little hard to understand. I think they would suspect there has to be some kind of good description of how we engage in reasoning and interpretation. So what does it mean for these things not to be formalizable?

Putnam: Well, take the case of whether the intuitive notion of proof, not proof in a particular formal system, but the intuitive notion of a correct mathematical proof, is formalizable. There is a sense in which that’s not formalizable, in the sense that if there were an algorithm which generated all and only correct mathematical proofs, we could not prove that it was correct [laughs]. And I believe the same thing stands to inductive logic. If you could formalize degree of confirmation, as Carnap hoped—if there were a program which assigned an “accept” or degree of confirmation greater than 0.5 to all and only those hypotheses it is rational to accept, then the hypothesis that that program, or any particular program, does that would itself be not one that we could confirm. It can happen that this is the great surprise of the Gödel theorems—that some of our abilities are such that we can partly formalize them, but with respect to any partial formalization, we are able to prove that it is only partial.

However, I agree that I was too hard on cognitive scientists in some of my writing. In “Cognitive Science or Science Fiction?” I did suggest that all cognitive scientists are basically computer nerds, that they’re all engaged in writing programs and in trying to reproduce our mental abilities entirely by writing programs, and I now think that was wrong. There’s a lot of cognitive science out there that is not that reductionist.

HRP: So then in what direction would you like to see the program of cognitive science take?

Putnam: Well, obviously “let a thousand flowers bloom.” I think there’s a lot of good cognitive science. The work of Amos Tversky, and people who followed him, like Maya Bar-Hillel, is also cognitive science.  This is an attempt to really describe what you might call useful errors in reasoning—including errors we had better make, because there isn’t always the time in the real world to use the best algorithm. I find that whole field fascinating. There is also the sort of thing that Ruth Millikan talks about—which again, I think I was too hard on in Renewing Philosophy—the idea of “normal biological functions,” which is a side of cognitive function and cognitive science that connects with evolutionary theory. In general, I think that our mental states are functional states—functional states not in the sense of computer programs—but they are states which have functions. I would call them functional states with long arms [laughs]. Basically that’s what externalism says: knowing the meaning of a word is having a long-armed ability, an ability that reaches out and can only be described by also describing the normal environment it’s geared to and what it responds to in that environment.

HRP: So, to that extent, evolution can explain representation—pace the title of one of the chapters in Renewing Philosophy.

Putnam: Well, let’s say evolution is certainly part of the explanation, but obviously most of it is cultural.

HRP: I’d like to turn back to realism for the moment. You discussed why internal realism failed, but I’d like to ask you about your current views on realism. What does that view look like?

Putnam: I would say that I am, in some sense of metaphysics, a realist in my metaphysics, although not a metaphysical realist in the overly narrow and overly specific sense that I attacked in my internal realist period. In my internal realist period, I saddled the metaphysical realist with having to reject conceptual pluralism and conceptual relativity—and of course if you do that, then you don’t really need the verificationism. As long as you recognize conceptual relativity and conceptual pluralism as real phenomena, then you’re already defeating metaphysical realism. But back then, I used to smuggle the verificationism in under the umbrella, but I would not do that now.

I would say yes, I am a realist in my metaphysics.  I would also say I am a scientific realist; scientific realism is a part of my realism, but have never thought of it as the same as metaphysical realism. I think scientific realism is something that I believed in during my internal realist period, and before and after, because scientific realism deals with the issue of whether scientific theories are just computing devices or just prediction devices, and says, “No, they are descriptions of reality.” And you can be a metaphysical realism and reject scientific realism—Bas van Fraassen is my example. And on the other hand, you can be a scientific realist and reject metaphysical realism, as I did in my internal realist period. I believe my “no miracles argument” [in Meaning and the Moral Sciences] is correct—that is, I believe the argument that positivism and instrumentalism make the success of science a miracle, and that can’t be right. And for me that’s still a cornerstone.

HRP: You said at a recent lecture at Boston University that scientific realism need not be entity realism. What does such a scientific realism look like?

Putnam: Well, remember Wittgenstein’s Tractatus. He said the world consists of facts, and not of things. I think there are facts, more broadly states of affairs, which are mind-independent, but I think the particular language we use, not just the natural language we use (although that’s a big factor) but also the scientific language we make up, often provides different descriptions of the same state of affairs, and very often those descriptions don’t have the same ontology or ideology in Quine’s sense. I used an example [in that lecture] from quantum mechanics, cases where what physicists regard as one and the same state of affairs, can be described in terms of fermions or in terms of bosons. This is a very startling and striking example, called duality in modern quantum physics, and there are lots of examples of duality—in fact, it’s one of the big themes in string theory. This is a phenomenon that I call conceptual equivalence.  So basically here I would say duality seems to be compatible more with Tractarian realism than with Quinean realism—not realism about entities but realism about states of affairs.

HRP: So then there’s no tension conceptual relativity and pluralism on the one hand and scientific realism on the other?

Putnam: That’s what I think, yes.

HRP: Again, I imagine that view is motivated by the success of science.

Putnam: The success of science, yes. The fact is that these cases of conceptual relativity are cases where the theories are actually intertranslatable. These aren’t translations that a linguist would regard as correct, but why should a physicists care? [laughs]

HRP: So we have a harmony between conceptual pluralism and scientific realism, and in your discussions about speaking objectively and having objective knowledge, you’ve spoken about “warranted assertability.” Could you describe what it is for an assertion to be warranted?

Putnam: Well, for an assertion to be warranted is for it to be reasonable to accept it. And of course, Tim Scanlon, whom I assume you have covered in the magazine or will cover in the magazine (I’m referring to the Locke Lectures he recently gave) defends what he calls “reasons fundamentalism,” by which he means the idea that reasonableness is not reducible to anything else, but nevertheless we should not conclude that it is all just culturally relative or, like the late Richard Rorty, what your cultural peers like to believe, etcetera. And I’m a reasons fundamentalist in Scanlon’s sense—once you go soft on the notion of reasonableness, then of course you’ve automatically given up scientific realism; as far as I can see you’ve given up every form of reasonableness and that does not seem reasonable to me [laughs].

HRP: But certainly how we cash out the notion of reasonable here is not supposed to be, say, phenomenological—that is, something is reasonable if it just seems reasonable. Can you describe what reasonable is in this sense? You can’t formalize it, certainly. . . . 

Putnam: It’s not just that you can’t formalize it, but we have criteria for being reasonable and unreasonable, and I think the world itself forces us to revise those criteria from time to time. That’s one of the reasons it defeats formalization. And even one of the great positivists, Comte—who, I guess, gave us the name “positivism”—said there is no such thing as the scientific method. Each science has its own method. Certainly he didn’t mean there’s nothing in common—all the sciences owe something to Bacon. They all recognize the idea of putting questions to nature, they all share the Baconian rejection of the idea that science can be done a priori. We replaced the idea of asking what Aristotle said by asking what nature says and putting questions to nature.

Of course, Habermas introduced the idea of discourse ethics, but I think you can find all the principles of Habermas’ discourse ethics in Dewey. I once taught a seminar with Seyla Benhabib that turned into a sort of Dewey versus Habermas seminar. She was arguing for the superiority of Habermas over Dewey, and I pulled out things that Dewey wrote in the thirties, things that contain explicitly every one of the principles of discourse ethics that Habermas listed [laughs]. You know, principles like people must have the right to make objections, ask questions, freedom of thought, freedom of inquiry, taking objections seriously, etcetera. These things are all that I think Dewey meant by “being scientific,” and so any serious field of scholarship should aim at being scientific in that wide sense.

But in the more narrow sense, I think you learn what it is to proceed scientifically if you’re going to be a molecular biologist, in your first couple of years of graduate school; and you learn what it is, in say string theory, in your first couple years of graduate school. And what the string theorist learns and what the molecular biologist learns are not the same thing. And in fact, if you believe Peter Galison, what the experimental physicist and the theoretical physicist learn are not the same thing—that’s why they have to use what he calls “local pidgens” and “creoles” to communicate with one another. Wittgenstein famously said “mathematics is a motely” but it is much more true that science, as a whole, is a motely.

But there’s something else that keeps us honest besides discourse ethics. Of course discourse ethics is a part of objectivity—and according to Rorty it was all of it; according to Rorty, all there is to the notion of objectivity, at least in our Western culture, which is the one [Rorty] liked, are these principles of discourse ethics. And he was a great admirer of Habermas, paradoxically. [Rorty] thought “Yes I can just talk about discourse. I don’t have to talk about the world.”

But it’s not just methodology—you know, the methodology of let’s say sophisticated empiricism—which I think was Newton’s correction to Bacon (that’s another subject), but I think Newton saw Bacon as putting too much weight on direct observation, and that what you have to do is combine the notion of induction—Newton does use the word induction—with mathematical analysis. And Newton was the first to see that there could be a mathematically sophisticated empiricism. If Dewey had known enough mathematics to really understand Newton’s Principia, and not just admire it, I wonder whether the history of pragmatism would have been different [laughs]. So I think there is something we can say about the methodology of objectivity, and it has been said.

But also we have to say that scientific methodology itself is something that the world sometimes forces us to correct. I think that in the case of quantum mechanics, quantum mechanics violates—I hope it won’t always—a lot of what people in the nineteenth century thought and had good reason to think was required of a scientific theory. One thing that was required of a scientific theory was that you should be able to understand it. And of course the question of how you could understand quantum mechanics is controversial to this minute. So we had to widen our concept of what’s even allowable in theoretical physics. And that’s a manner of being sensitive to the demands of nature itself.

HRP: To follow up on that a little, in a lot of your writings you talk about things such as principles of charity, principles of economy, or cases such as choosing between Einstein’s relativity or Whitehead’s alternative theory of gravitation—or in this case, choosing to keep the reference of terms like “momentum” the same across theories. Are these what you mean by “constitutive standards of rationality”? 

Putnam: These are provisionally constitutive. Nature may force us to go back and rethink. I don’t believe in any absolute a priori. But these are provisionally a priori principles of scientific procedure.

HRP: You mentioned Habermas, and it seems, at least, that as your career has progressed, you have grown more and more interested in Continental thinkers such as Derrida, Foucault, Habermas, and the like. Can you describe the advent of this interest and what you get from these thinkers?

Putnam: When I first hear that question, I think “That would have been true in 1990,” but it hasn’t really been true since. My interest in the French intellectuals has somewhat declined, with one big exception, but he’s not a postmodernist or a deconstructionist—Pierre Hadot—who died a couple of weeks ago. In fact, we corresponded just before he died. A “Philosophy of Hilary Putnam” volume is going to appear in the Library of Living Philosophers, and Hadot has a paper in that—and I had just sent him my very approving reply to his paper, and we exchanged emails about that, just before he died. But Pierre Hadot is not a deconstructionist—in fact, he was the first French philosopher to write on Wittgenstein.

Really, my interest in Derrida very much declined after 1990. I wrote a critical chapter on him in Renewing Philosophy, and I do think some of his ideas are interesting. For instance, asking whether a philosopher’s tropes are really compatible with what appears to be the literal meaning of his theses is a very good idea. A particular example: one might ask in the case of Quine, when he says that “reference floats,” when he employs the trope of reference floating, and says that this is all right as long as the boats stay together—well, in that trope, a good Derridean might point out that it looks as if the water has to be something like a noumenal reality [laughs]. The trope is totally incompatible with Quine’s philosophy. I think in general looking closely at the literary devices a philosopher uses and asking how compatible they are with what seems to be the literal meaning of the theses he defends is a good idea.

But nevertheless, there’s an awful lot of gobbledygook.  I mean, the idea that we discovered in France that there’s no such thing as truth, and no such thing as the real world—this is the way that English departments and French departments expressed their enthusiasm for Derrida, as if we had discovered that there is no such thing as objectivity—is absurd. Even Derrida stayed short of saying that there’s no such thing as truth, but he sometimes suggested that if we were daring enough, we would go that far, but for some strange reason we can’t. So I don’t find very much I can use in Derrida.

Foucault as well goes too far sometimes, in my view, but Ian Hacking and a former student of mine—Arnold Davidson at Chicago—have gotten something important from Foucault, namely the idea that there’s no contradiction between analytic philosophy, that is analyzing concepts, and historical analysis. The fact that you’re interested in the analysis of concepts doesn’t mean that you’re committed to a kind of Platonism in which you treat concepts as the sort of things that have no history.

In particular, Foucault and Hacking and Arnold Davidson are right that the concepts in the human sciences used to describe people may actually change the way people behave. Consider the way the concept of homosexuality succeeded the medieval concept of perversion, an example that Arnold Davidson has used: there was no “medical problem” of homosexuality for the medievals—“perversion” then was just a sin. And then suddenly, when it became a medical category, first people tried to explain homosexuality as a sign of genital malformation (!), but it turned out they couldn’t find a genital malformation, so they tried to find a brain malformation. And psychoanalysis succeeded that, with the idea that homosexuality is a “neurotic symptom”, and after that (and fortunately) we realized that the assumption that there is something wrong with homosexuality is what needed to be challenged.  So in these areas, doing Foucaultian analysis, with perhaps more responsibility to the facts than Foucault had, and doing conceptual analysis are not incompatible, and they shed light on one another.

HRP: Something that seems to emerge later in your writings is an interest in ethics. Can you describe the relationship between your other work and your ethical philosophy?

Putnam: I tried in my last book, Ethics without Ontology, to argue that if you step back, you can see that many of the puzzles raised about how ethics can be objective have exactly the same form as puzzles raised about how mathematics can be objective. And when you see virtually the same arguments being applied to subject matters as different as ethics and mathematics, then asking “What picture are these philosophers in the grip of?” is a good idea. And here I think it is something like a picture that objectivity requires objects—it’s almost built into the nature of the word ‘objectivity.’ You might think there must be funny ethical properties to make true ethical statements, or there must be funny objects—Quine once said that numbers are intangible objects—to make mathematics true. I think that’s a mistake. I think ethical judgments are judgments of what is required by certain fundamental human interests, and that ethics has a history. My problem with evolutionary ethics is that it confuses explaining altruism with explaining ethics, which is curious, because “ethics” is a Greek word, but “altruism” is a notion that doesn’t appear in Aristotle’s Niccomacean Ethics at all; there is no general virtue of altruism, and certainly not altruism in the sense of respecting the needs of the other regardless of social class or ability. Something like sociobiology can certainly shed some interesting light on how we’re able to be ethical, but again, that’s not all there is to ethics.

Historically, we have I think at least two huge competing ethical systems—one I call “macho ethics.” And fascism was in many ways a conscious intellectual revival of macho ethics, via a misreading of Nietzsche, whom I would not blame for it, but fascist intellectuals basically erected an ethical system based on the importance of virility, courage, daring, etcetera. Perhaps you could give an evolutionary explanation of our attraction to macho ethics, and you could give an evolutionary explanation of our attraction to compassionate ethics—but those causal explanations are not going to help you decide between them.

HRP: You say there don’t have to be ethical objects to talk objectively about ethics, and I take it that what makes our ethics objective is what we were talking about earlier—warranted assertability and what’s rational to accept.

Putnam: What’s rational to accept, given certain interests. And I think those interests, including the interest in human flourishing, the Scanlonian interest in giving one another reasons, and the interest in equality, generally support one another. There are of course cases where they conflict. The reason that they become a part of a single ethical package is partly historical. The interest in equality becomes a more powerful interest when we begin to have constitutional democracies. It’s no accident that equality becomes thematized so much at the time of Kant—this is precisely the time that constitutional democracy (or at least constitutional monarchy) is on the historical agenda. But I don’t think that fact entails cultural relativism, because science also has a cultural history, but that doesn’t diminish its objectivity.  I’m not a utilitarian, but most people are happier not living in a tyranny; and utilitarianism is a quick and dirty—too quick and too dirty—way of responding to the question “which makes one way of living better than another?”

HRP: You said that what’s rational to accept is conditional upon certain interests, but I suppose you don’t want to say that one of them will trump the others. They’re pluralistic on a plane, or something like that.

Putnam: They’re so interlocked. It’s so easy to write a book deriving n-1 of them from the nth [laughs]. And of course that game has been played and will go on being played. I prefer to see each of them having a somewhat different history, but because they support one another, and especially in conditions of democracy and mass public education, that they have come to be virtually a single package. But—and this is a Habermasian point—it’s important that when people were denied some of them in the past, people had to come up with empirically false claims to support denying them. So it’s not simply arbitrary. Thus, for example, the long, long inequality of women was justified by absurd claims, such as Aristotle’s claim that women don’t have active nous, the highest, intellectual form of nous.  Similarly, racial oppression was justified by false claims.

HRP: So that’s what happens when they jive, but when they conflict, let’s say, do you have a pragmatic maxim for how those kinds of conflicts should be resolved?

Putnam: Try to be intelligent [laughs]. No, there is no decision algorithm. That’s something Kant said: there is no rule which will not give bad results if conjoined with universal idiocy. The one thing there’s no rule for is good judgment. Here he uses a nice feminist term: “mother wit.”

HRP: I’d hate to ask an overly general question, and it might be unfair, but I think people would appreciate hearing an answer from such a distinguished philosopher. Where do you think philosophy is going in the current century, and where do you think it should go?

Putnam: One theme in my work is the idea of a philosophy that respects science and takes science seriously, but doesn’t have a positivist or otherwise ideological picture of what science is, and that doesn’t think that philosophy can be reduced to science.  Obviously that’s the way I think philosophy should go, and I’m optimistic that there are probably more people who think that way now than there were when I started. The other main theme is that objectivity, truth, and warranted assertability are not restricted to science; values judgments and descriptions of facts are “entangled,” to use a term I learned from John McDowell, and the objectivity of science is connected with and presupposes the objectivity of values.

HRP: Can you speak more about this distinction between a scientistic view, or an ideologically motivated conception of science, as opposed to science itself?  I take it you have in mind something like Bernard Williams’ “absolute conception of the world.”

Putnam: Well, I also have the positivists in mind. I was just re-reading the first volume of the Minnesota Studies in the Philosophy of Science. There’s a long essay at the beginning by Herbert Feigl, in which he says that positivist philosophy of science has become more realistic, but we have to insist insist on the sharp line between the analytic and the synthetic, we have to say the meaning of a term is the rules of its use, and so on. There was a very strong feeling among the positivists that they had told us exactly what science is—they had married Bertrand Russell’s romance with mathematical logic with classical empiricism. That’s, in the wide sense, an ideologically driven concept of science.

Bernard Williams was a physicalist who knew no physics. I loved him, and he was a brilliant philosopher, but his admiration for physicalism was in inverse proportion to his actual knowledge of the physical sciences.

HRP: But it seems surely no one anymore is a positivist in that sense.

Putnam: I suppose you could find one somewhere [laughs].

HRP: So do you think this ideologically driven notion of science has dropped out, or do you think it’s taken new form?

Putnam: Well, it may reappear. It may certainly reappear. It’s always premature to say that a mistake in philosophy has been defeated once and for all. If someone wants to try and show that the program can be carried through, they should try. We learn from our failures in philosophy. Right now, I would say one of the greatest dangers is reductionism, which often blinds people from considering more sophisticated alternatives—take evolutionary psychology, for instance. I don’t think that it’s illegitimate, and I think there’s a lot to be learned from it, especially when the evolutionary psychologist, as it happens in the best cases, can suggest consequences of his theory that can be tested now. Those are the exciting cases. But in the worst case, it becomes Just-So Stories. And again, the problem with AI  [“Artificial Intelligence”] was that there was so much oversell. There we tremendously exaggerated claims both about what had been achieved, and what would be achieved in the next five or ten years.

HRP: What do you have in mind when you say “sophisticated alternatives to reductionism”?

Putnam: In general, one can study how A influences B, or how A depends on B, without thinking that that’s the whole story.

HRP: So essentially you mean tampering the stringent requirements of classical reductionism—bridge laws, lawful covariance, and such?

Putnam: That’s right.

HRP: One picture that emerges from this view of reductionism, and indeed the foregoing conversation, is a skepticism about theories of meaning and theories of intelligence, for all the reasons we’ve discussed. Jerry Fodor put it in a review that he took you to be denying that meaning is a natural kind. Is that a fair characterization of your views?

Putnam: No, I don’t think I’ve ever said that. I think meaning might be, in a sense, a natural kind—in the way that biological kinds are natural but don’t have sharp boundaries. This is what Ernst Mayr taught us all: yes, species are natural kinds, but it’s sometimes arbitrary when you say here is one species and here are two. If meanings are natural kinds, they’re much more like biological kinds than, say, chemical kinds: they’re historical natural kinds. Now my old friend Noam Chomsky—we’ve been friends since college—has written that there are no such things as languages, for example, because there are only criteria for idiolects but not languages. And I would say, sure, it’s arbitrary where you put the precise borders of the French language, for example. There are so many dialects of Dutch, and yeah, there’s not a sharp line there. But it’s really scientism in the bad sense to say we can only study what can be mathematized—we’ll only look for algorithms to describe what goes on in the brain when it uses certain idiolects. In fact, Noam even rejects talk of reference, because it’s a hopelessly unscientific notion, so just forget it. Well, no! The world won’t let us! [laughs] It’s very important to us what people are referring to when they use a term. So I would say that it’s not that there are no natural kinds here, but that explanations have to seek their own level. The standards of explanation of one science can’t just be simply transported to a different science.

I think there’s a lot to be done in the study of meaning. My own prejudices are that I still think that lexical semantics is sadly neglected. There are some very good exceptions—mostly on the West Coast. There’s something about the East Coast, maybe it’s too close to MIT [laughs], that they talk less about the meanings of words. But I think it’s very important to study the meanings of words.

HRP: That’s an interesting  division between you and, say, Davidsonian  semantics. Why such an interest in lexical semantics?

Putnam: Well, that’s how you learn a language, to begin with. You learn to use words!

HRP: You’ve written on Jewish philosophy recently. What do you take from it, and what’s the interest?

Putnam: I think for me, as a religious person and as a naturalist, that means that Judaism can be a form of spirituality—but not necessarily a theory about immaterial, supernatural objects. And spirituality interests me both because I think I experience a form of spirituality that the Jewish tradition makes available to me, and I don’t think that spirituality is something that’s going to go away. I think Richard Dawkins has this simplistic division of the world into fundamentalists, who believe in the literal truth of their local myth, and smart people like him who are hard, scientific types [laughs]. And recently, he started admitting that there are more liberal types—but that seems to be a late afterthought. But that’s just not the way human life is. I think religion can either make you do terrible things. Dawkins was right about that, although I’m not sure at all that, say, the Thirty Years’ War was really a war fought between Catholic states and Protestant states—how come Catholic states were on opposite sides of the Thirty Years’ War, and how come the main Catholic general was a Protestant from Sweden? You could probably make a stronger case that it was an imperialist war, and you had to tell your local population something about it, so you invoke the church. Just as we invoke democracy with equal fraudulence.  But religion has led to fanaticism and other evils at times, without question.

My affection for Jewish philosophy fits in, I suppose, with my general sense that we can’t make everything scientific—that philosophy is closely related to the sciences, but philosophy also has a humanistic side. And I have long felt, as Kant did—this is something that the editors of one of my Harvard volumes pointed out—that philosophy needs both a humanistic conception of what it’s doing and a scientific conception of what it’s doing. It’s always been the case that what’s lasting in philosophy arises from the interaction between the scientific side of philosophy and the humanistic and ethical side of philosophy.

Regarding the part of philosophy that is more than just science, I don’t think it can or should hope to find final answers to age-old questions, or even to remain content with those questions as they were traditionally formulated, but I think there are certain questions in ancient and medieval and early modern philosophy that we can recognize as being continuous with questions we’re asking today—for example, why is there a difference between right and wrong? How come there is such a thing as mathematical truth? These are certain questions that won’t go away. Although their verbal formulation may go away, something like them always comes up in their place.

1. INTRODUCTORY REMARKS

My goal here is two-fold. First, I want to accurately describe as best I can the actual methodology employed by analytical epistemologists, at least as I understand that methodology. In my view there has been a tendency in the recent debate about the methodology of analytical philosophy to offer a far too crude account of what it is that analytical philosophers do when they do philosophy, and this is especially so when it comes to describing what analytical epistemologists do. Second, I want to defend this methodology from some recent attacks in the literature, especially from exponents of a particularly robust form of experimental philosophy. In particular, I claim that once we understand the methodology actually employed by analytical epistemologists properly then we will see that it is not as exposed to certain challenges in the way that some have claimed.

In order to keep my discussion focussed, I will train my attention on the specific epistemological task of offering a theory of propositional knowledge.

2. EXTENSIONAL, INTENSIONAL AND GENERAL INTUITIONS

Contemporary epistemologists engaged in the project of offering a theory of knowledge are often characterised as appealing to just one set of data as ‘input’ to that theoryviz., the data provided by their intuitions about knowledge. As we will see below, I think this way of describing the methodology of contemporary epistemology is far too simplistic. But even insofar as we focus on the philosophical data provided by intuition, I still think we need to recognise that this data comes in various categories, some of which carry very different epistemic weight when compared to others.

Much of the focus when it comes to the role of intuition in epistemology is on our intuitive responses to cases, where we are asked to form an intuitive judgement about whether the target term is applicable in the case under discussion (which may be actual or hypothetical). Call such intuitive judgements, extensional intuitions. As regards knowledge, the obvious example to give in this regard is the famous Gettier-style cases in which we are offered situations in which an agent has a justified true belief and yet their belief is only true as a matter of luck (cf. Gettier 1963). Such cases are held to elicit the intuition that the agent lacks knowledge, and thus that the extension of this term does not cover such cases. This in turn gives us reason to believe that knowledge is not simply justified true belief.

There is much more to the role of intuition in epistemology than extensional intuitions, however. In particular, one key source of intuitional input to epistemology is that provided by our intuitions regarding the intension of the target term. Call these intensional intuitions. With regard to the specific epistemological project that concerns us, the relevant intensional intuitions will be about knowledge.
Here are some claims about knowledge which might plausibly be offered as intensional intuitions:

(i) Knowledge that p entails p;
(ii) S’s knowledge that p entails that S believes that p;
(iii) S’s knowledge that p entails that S is in possession of reasons for thinking that p is true;
(iv) S’s knowledge that p entails that S’s belief that p is not true simply as a matter of luck;
(v) S’s knowledge that p is the result of S’s exercise of a relevant cognitive ability.

This list is clearly not exhaustive, and nor are all the claims on this list beyond contention. But insofar as these claims are intuitively true, then they appear to specifically constitute intuitions regarding the intension of the target term.

Now clearly there will be an interplay between our extensional and intensional intuitions about a term. Perhaps, for example, one’s inclination to judge that knowledge is factive is a reflection of our inclination to judge that agents lack knowledge in cases where the proposition believed by the agent is false. Alternatively, perhaps the direction of fit is the other way around, and our extensional intuitions in this case reflect a prior intentional intuition about the factivity of knowledge. But even though there might be close connections between the relevant extensional and intensional intuitions when it comes to specific cases, we clearly cannot take it as given that this is so across the board, particularly since there seem to be cases where our extensional and intensional intuitions are in conflict (we will consider some examples presently).

There is a further type of intuitive data which is relevant here, a kind of intuitive data which is closely related to intensional intuitions but ultimately distinct. For in addition to having intuitions about the intension of a term, we can also have general intuitions about that term which are neither about its intension or its extension. For example, that knowledge is more valuable than mere true belief is often cited as an important epistemological intuition, but while this intuition is clearly about knowledge, it does not fall into either of the two categories we have demarcated. Other intuitions of this sort could be that knowledge is an ‘inquiry-stopper’, or that knowledge is the norm for assertion. Call this class of intuitions general intuitions.

Within any one domain, such as epistemology, there will inevitably be a complex interplay between these three categories of intuition. One might, for instance, have an intuition about the intension of a term which is called into question by an extensional intuition one has regarding a case in which the term has application (or lacks application). This will obviously generate a theoretical tension that will need to be resolved, perhaps by highlighting an important ambiguity in the intensional intuition or by showing that the extensional intuition can be explained away.

Similarly, one might be led by one’s intensional and extensional intuitions towards a theoretical account of a term which is in conflict with one or more of the general intuitions. For example, one’s intensional and extensional intuitions might lead one to adopt a particular account of knowledge, but one might then realise that this account is unable to accommodate an important general intuition about knowledge, such as the putative general intuition noted above that knowledge is more valuable than mere true belief. Again, this will result in a theoretical tension that needs to be resolved. And of course, even within a particular category of intuition one could have intuitions that conflicte.g., by having extensional intuitions which seemingly support different theories of knowledge.

The foregoing reminds us that intuitions are not set in stone but are instead highly defeasible, even when we consider them only in the light of other intuitions (we will consider some non-intuitive data that is part of the epistemological project in a moment). This raises the interesting question of just how defeasible our intuitions are, which brings me to a further point. For while I think we should regard all of our intuitions as individually defeasible, I also think we should resist putting all our intuitions on a theoretical par. Indeed, I want to suggest that some intuitions will be less defeasible than others, and some classes of intuition may well be collectively indefeasible.

As an illustration of this point, notice that our most deep-seated intensional intuitions about a term play a very specific role in the analytical project. This is because an account of, say, knowledge which did not respect many of our most fundamental intensional intuitions about this term would be unlikely to count as a theory of knowledge at all. Suppose, for example, that we offered a theory of knowledge on which it wasn’t factive, didn’t entail belief, didn’t require supporting evidence, was consistent with lucky true belief, didn’t require the exercise of a relevant cognitive ability, and so on. What would make such a theory a theory of knowledge, specifically? The point I am getting at here is that our most basic intensional intuitions about a concept play the role of picking out the very thing that we are trying to understand, and hence we cannot depart too far from them without losing that which we seek.

Call our most deep-seated intensional intuitions about a concept, intensional platitudes. The foregoing is consistent with us rejecting or revising some of the intensional platitudes as part of pursuing the analytical project, but not with rejecting or revising the intensional platitudes en masse. This is the very specific sense in which intuitions can have a kind of indefeasible standing within the analytical project.

This point about intensional platitudes ensures that they enjoy an enhanced epistemic role in the analytical project, since the process of reflective equilibrium undertaken by the analytical philosopher to square all the intuitive data (and much more besides, as we will see below) will need to pay special respect to these intuitions. Our intuitions can differ in epistemic standing in other ways too. Take extensional intuitions, for example. Some extensional intuitions can be extremely robust and thus carry a great deal of evidential weight. The Gettier-style cases are an obvious case in point here, though it is important to remember that even here the intuitions in question are still defeasible. But other cases might be far less compelling, even though they do still elicit the relevant intuition.

The extent to which an extensional intuition carries epistemic weight can depend on other factors too, such as the degree to which it trades on a real-life example. This is partly why I think the Gettier case is so compelling, in that it involves a very simple everyday scenario. Indeed, Gettier-style cases occur all the time in real life. In contrast, some extensional intuitions trade on examples which concern far-fetched scenarios (e.g., cases that appeal to science fiction), and some examples concern scenarios which may not even be possible, in some suitably robust sense of ‘possible’ (e.g., metaphysically possible). Ceteris paribus, the more far-fetched the example is, and particularly the more dubious the example is in terms of its possibility, the less epistemic weight any extensional intuition based on this case will have, for the simple reason that there is more scope to reject the example.

3. VIRTUOUS INTUITION

We have thus delineated three categories of intuition, and also shown how there is a complex interplay between them, both in terms of determining their relationships to each other, and in determining their respective epistemic weights. This complexity demonstrates the need for analytical philosophers to have a great deal of critical skill if they are to effectively make use of this intuitive data as part of forming a particular theory, such as a theory of knowledge. This is not the only way in which skill is required, however, since I think it is clear that it is also required in order to properly elicit the relevant intuitional data in the first place.

We can explore this point by delineating a further kind of category of intuitions which concerns intuitions about the correct linguistic usage of the target term. Call this category of intuitions, linguistic intuitions. So, for example, epistemologists might appeal to intuitions we have about the correct usage of the word ‘knows’ in a particular conversational context as part of the relevant evidence when formulating a theory of knowledge.

There is, of course, a close relationship between extensional intuitions about a certain term and linguistic intuitions. Nonetheless, it is still important to distinguish them, for although there will clearly be a great deal of overlap in the two kinds of intuitive judgements one makes about a particular term, such judgements can come apart, in both directions. The case of knowledge (/‘knowledge’) is no exception.

On the one hand, there can be cases where we would intuitively judge that an agent lacks knowledge even while granting that it would be in some sense legitimate for the agent to say that she knows. Gettier-style examples are a case in point here, in that given the rational basis the agent possesses in support of her belief it clearly would be appropriate for her to assert that she has knowledge, even though she in fact fails to know.

On the other hand, there can be cases where we would intuitively judge that an agent possesses knowledge and yet nonetheless hold that for independent reasons it would not be appropriate for her to (flatly) assert that she knows (even though this assertion would be of a truth). An obvious example that springs to mind on this score is a case where such an assertion, though true, would in that conversational context generate a false conversational implicature.

We have then a subtle distinction between extensional and linguistic intuitions, and now that this distinction is on the table it ought to be clear that part of the skill of the analytical epistemologist will lie in determining which specific intuition is being elicited in a particular case. For what at first appears to be an extensional intuition could prove to in fact be a linguistic intuition, and vice versa.

This is just one way in which intellectual skillwhat we will refer to as ‘intellectual virtue’is required when undertaking the analytical project. A badly constructed example could obscure things, and so it may take a very carefully constructed case in order to draw out the relevant extensional intuition. A poorly formulated intensional intuition might gloss over an important distinction, and therefore be very misleading. A failure to attend to the ways in which the explicit usage of ‘knows’ within an example can affect our intuition about whether an agent in an example has knowledge could muddy the philosophical waters. And so on.

The intellectual virtues on display here—close attention to the salient details, a keen eye for spotting potential ambiguities in cases, and so on—will also be on display when the analytical philosopher adroitly conducts the complex process of reflective equilibrium that is required in order to ‘weigh-up’ all the relevant inputs to the analytical project. As noted above, this process is complex enough even if one restricts one’s attention only to the intuitional input. Since, as we will see below, the input to the analytical project goes beyond intuitions, the process is even more complex still.

That it is part of the methodology of analytical epistemology to regard intellectual virtue as being required in order to properly attend to the intuitional data is clear from considering how analytical epistemologists respond to intuitions. Indeed, the very fact that Gettier-style cases, when first presented, were able to shock epistemologists into seeing that the received wisdom in the theory of knowledge was wrong demonstrates that great skill can sometimes be required in order to elicit the relevant intuitional data. Because intuitions are most naturally understood as a kind of ‘intellectual seeing’, it is tempting to think that the intuitive data is manifest for all to see, but a moment’s reflection reveals that this would be a non-sequitor. For as the Gettier-style cases reveal, it can in fact be quite hard to draw out this intuitive data.

Further support for the claim that epistemology is wedded to the idea that intellectual virtue is required to properly elicit intuitions can be gleaned by examining how the intuitional judgements of subjects who have been given the relevant analytical skills are privileged over those same subjects’ first reactions to cases. Call these two types of intuitional judgement, initial intuitions and intuitions under reflection (where, note, the latter are still bona fide intuitional judgements, and hence are non-inferential).
To illustrate this contrast, consider first a case outside of epistemology, that of Robert Nozick’s (1974) ‘experience machine’. This is a machine that creates an artificial life for the subject experientially indistinguishable from ‘real’ life, in the sense that once one is in the machine one can’t tell that one’s experiences are in fact artificially generated. Let us stipulate that life inside the machine is significantly more pleasurable than normal life outside the machine. Here is the philosophical question: should one prefer an artificial life inside the machine, with all its additional attendant pleasure, to a real life outside the machine with all its attendant trials and tribulations?

My experience as someone who has often taught this example to students who are encountering philosophy for the first time is that insofar as the students have any initial opinions on this matter at all, then they tend to intuitively regard the life in the experience machine as at least no worse than the real life, and often preferable to the real life. Significantly, however, this judgement tends not to be stable. For example, if one asks the students whether they would be happy for their children to live their lives in the experience machine then most opt for the real life outside of the machine, even though they recognise that there is a tension between this judgement and the previous judgement about the desirability of the life in the machine. Relatedly, if one makes explicit that entering the machine is a one-way ticket—perhaps because one’s body becomes unusable thereafter as part of the ‘re-orientation’ process—then again students’ intuitions tend to shift towards regarding the life outside the machine as being preferable to the life inside the machine. In fact, once one has explored the example in some detail then the groundswell of opinion tends to be in favour of treating the real life outside the machine as better than the artificial life inside the machine.

Here, then, we have a case in which people’s initial verdicts about a scenario change over time in response to questioning and further reflection. Similar cases can be found in epistemology. For example, in my experience quite a few students introduced to epistemology for the first time are inclined to hold that there can be knowledge of falsehoods, though this judgement fades under questioning and further reflection.

A model of intuition which treats intellectual virtue as a requirement to best elicit intuitions can explain what is going on here. The process of reflecting on the example and identifying salient details is in effect training the subject to apply the relevant intellectual virtues so as to improve her ability to discern the relevant intuitional data. What we are ultimately seeking from subjects is thus not their initial intuitions at all, but rather their intuitions under reflection, where the latter (if we have done our job well enough anyway) are the product of the application of the relevant intellectual virtues.

An interesting issue regarding this account of the role of intuition in epistemology is what epistemic weight, if any, to attach to the initial intuitions. A hard-headed response would be to say ‘nothing at all’, on the basis that the corresponding intuitions under reflection effectively ‘trump’ them. I think this would be a mistake, however. In particular, I think that the fact that subjects’ are inclined to make the initial judgement retains some epistemic weight, and so has a part to play in the process of reflective equilibrium which is undertaken by the epistemologist.

Indeed, given that all intuition is defeasible, even when the product of the relevant intellectual virtues, it would be foolhardy to discount the epistemic weight of the initial intuitions altogether. For even if one has enough faith in the epistemic credentials of the analytical philosophical enterprise to suppose that those trained in this enterprise have greater skills on this score, one would still have to take into account the possibility that the ‘training’ had itself introduced error.

Now, one way in which error could be introduced via this process can be discounted—viz., where the agents concerned only come to form their ‘intuitive’ judgement after training because they are simply deferring to the testimony of the ‘expert’ in the room. It is undoubtedly the case that some people will be influenced this way, but it is irrelevant for our purposes since such judgements are not intuitions at all, on account of their being epistemically mediated.

Still, even setting aside potential error of this sort, there is nevertheless scope for the analytical enterprise to generate error in one’s intuitive judgements. For even if one holds that analytical philosophy can help an agent clarify the conceptual landscape, and thereby enable her to better form intuitive judgements, it is entirely consistent to also hold that this process can sometimes muddle the conceptual landscape, and lead to error. Philosophers, even philosophers who count as ‘good’ philosophers by analytical lights, have persuasively argued for theses that have been subsequently derided. Moreover, the fact that there is widespread disagreement in philosophy should surely give us pause when it comes to supposing that analytical philosophy always enhances one’s evidential position when it comes to forming intuitive judgements.

A final issue here is the potential for there to be a substantial conflict between the initial intuitions and the intuitions under reflection. It would be an odd epistemologist who was entirely unconcerned about the fact that her theory of knowledge directly conflicted with the folk’s initial intuitive judgements about knowledge. At the very least, that there is such a disparity calls out for an explanation.
4. NON-INTUITIONAL INPUTS TO A THEORY OF KNOWLEDGE

I maintain, then, that moving to a virtue-based model of intuition does not oblige one to dismiss the epistemic weight of subjects’ initial intuitions. Interestingly, however, when one takes into account the fact that subjects (including, possibly, one’s previous self) are inclined to make initial intuitive judgements of a certain sort, one is in effect appealing to empirical data. Moreover, although this data concerns subjects’ intuitions, it clearly provides a very different input to one’s theory to that provided by one’s own intuitions under reflection. To this extent, we can describe it as non-intuitional input to the theory of knowledge.

I don’t think there is anything puzzling about this. For notice that just as we accord subjects’ initial intuitions some epistemic weight when formulating a theory of knowledge, so we also accord some weight to the fact that other philosophers have different intuitions to oneself (even though they are also intuitions under reflection). For that too is a factor to take into account when offering a theory of knowledge, and, relatedly, something which calls out for explanation.

In any case, given that the initial judgements made by subjects constitute empirical data that is relevant to the epistemological enterprise, it follows that epistemologists have an interest in the careful and systematic collection of this data. Although epistemologists have a role to play in this regard, such a task is probably best led by other specialists who have the specific skills that are salient to this undertaking. There is, for example, a wealth of empirical data on the kinds of heuristics and biases that blur human reasoning, and which would undoubtedly play a role in generating some of the intuitive responses given by subjects. Knowledge of these heuristics and biases would therefore be essential to the optimal gathering of the relevant data.

A second way in which a more informed and systematic collection of this data would be useful is in looking for whether some of the judgements about the target term are less universal than others. ‘Knows’ is a universal lexeme, which means that it is one of the few terms that appears in all known languages. Even so, there might well be widespread divergence in the intuitive judgements offered with regard to this term which reflects particular racial, cultural, social (etc.,) divisions. Given that many analytical philosophers get their data in this respect from a highly unrepresentative population sample (for the most part their undergraduate students), there is real potential for one to discover some very surprising empirical data on this score.

To this extent, I think that experimental philosophywhich aims to bring the specific expertise to bear in this regard in order to properly elicit this datais a welcome development. Indeed, given the fact that such data does have an epistemic weight within epistemology (and, for that matter, analytical philosophy), it is odd that there hasn’t been much interest until quite recently in appropriately collecting this data. Note, however, that this way of thinking about experimental philosophy sees it as merely assisting contemporary epistemology by helping it to collect better evidence of a certain type that is relevant to the project. It does not see experimental philosophy as posing fundamental problems for the epistemological enterprise, which is what some of its proponents claim. I will return to this issue below.

This is not the only kind of non-intuitional input which is relevant to offering a theory of knowledge. For example, one kind of additional data that epistemologists can appeal to takes the form of information about the genealogy and/or teleology of the concept in question. One aspect of this data is straightforwardly empirical. In learning something about how the concept in fact came about we thereby have information which can potentially play a role in helping us formulate a theory of knowledge. Of course, the epistemic significance of this data is not straightforward, and it is also highly defeasible, but as we saw above, that is also the case with the rest of the input to a theory of knowledge, with the exception of the intensional platitudes.

Note too that there is also potentially an a priori route to gaining an insight into the genealogy and/or teleology of a concept. A good example of how this might take place is offered in a seminal work by Edward Craig (1990). Here Craig asks us to imagine a ‘state of nature’ in which creatures with very similar interests and cognitive capacities to our own, and who occupy similar environments, nonetheless lack the concept of knowledge. The question that Craig asks is what would prompt agents in such circumstances to acquire this concept? Craig then offers us a complex, but highly plausible, answer to this question, one that informs his own theory of knowledge. The details of such a proposal are not relevant for our purposes; what is relevant is just that this account does seem able to offer us some input for our theory of knowledge (defeasible input, as always), and that the input in question is of an a priori nature, although it does not seem to fall into any of the categories of intuitive input noted above.
There will be other forms of data that epistemologists can appeal to in constructing a theory of knowledge—for example, relevant work done by cognitive scientists—but the foregoing should suffice to demonstrate the point that we are not limited to intuition alone.

5. THE METHODOLOGY AND GOAL OF THE THEORY OF KNOWLEDGE

I noted earlier that the methodology of analytical epistemology is essentially the application of a process of reflective equilibrium to the inputs just described. What the epistemologist is seeking is a reflectively stable theory of knowledge which can accommodate all these inputs. This is no easy task. As noted above, even within a single category of intuition-derived data there can be conflict, and the scope for conflict once all the varieties of inputboth intuitional and non-intuitionalare taken into account is huge. The task in hand for the epistemologist is to use her critical and logical acumen and, with a healthy dose of ingenuity, find the best way of squaring all this data.

The result may be quite revisionary, especially since, as we noted above, the only real constraint is that at least some of the intensional platitudes must be treated as sacrosanct. One would also expect there to be a fair degree of divergence amongst epistemologists in terms of the respective reflective equilibria that they reach. For one thing, attending to all the relevant evidence and ascribing it its due weight will be very tricky, especially since there is a feedback loop in play here, in that the weight one ascribes to one set of evidence will inevitably influence what weight one ascribes to other sets of evidence. I don’t think this should surprise or (in itself anyway) concern us, since it just reminds us that analytical philosophy is hard.

One issue that I have glossed over so far is what the epistemologist is trying to achieve with this methodology. There is a caricature of epistemology which has it as striving for an informative, but still fully reductive, analysis of knowledge, where this involves a ‘decomposition’ of the target term into its constituent conceptual parts without at any point appealing to the defined notion itself. Worse, it is sometimes alleged that the concept of knowledge being defined is to be identified with the ‘folk’ concept, such that where there is any significant mismatch between the theory and subjects’ initial judgements regarding the target term then the theory is potentially in jeopardy.

I don’t deny that it would be intellectually appealing to be able to offer a fully reductive theory of knowledge, but I think that it is a Herculean endeavour. Fortunately, the project of offering a theory of knowledge does not stand or fall by whether it can attain this end. In fact, I think that the actual practice of epistemologists in this regard—indeed, of philosophers more generally when offering a theory of a particular philosophical term—is more aimed at what P. F. Strawson (1992) referred to as an ‘elucidation’ of the target notion, where this involves an informative account of that notion which need not be fully exhaustive or reductive.

Two points are in order here. The first is to recognise that we are not simply trying to offer a theory of the folk notion. That much should be clear from what we have said in the previous sections. For while what the folk are inclined to say about knowledge (their initial intuitions) is of course data for the theory, such data is nonetheless defeasible, and needs to be considered in the light of other inputs into the theory, not least one’s intuitions under reflection. Of course, we do want our theory of knowledge to give special weight to the intensional platitudes, and one would expect these platitudes to reflect in a fundamental way the folk usage of ‘knowledge’, so to that extent one would not expect there to be too much of a discord between the theoretical account on offer and the folk usage. Nevertheless, given that any particular intensional platitude is not immune to revision this does not prevent there being a significant disparity between the folk usage and the theory account put forward. Indeed, we would reasonably expect the theory to depart from the folk usage of ‘knowledge’, not least because the folk usage does not itself suggest a consistent theory of this notion. The theoretical account we seek is thus likely to be a kind of ‘cleaned-up’ version of the folk notion, albeit an account which is essentially tied to the folk notion in virtue of the special weight that is attached to the intensional platitudes.
The second point is that a mere elucidation of a term can in fact give us just what we seek in terms of being a theory of that term, and can even be sometimes more informative than a reductive analysis. A good example in this regard is virtue epistemology. Suppose that it turns out that the best theory of knowledge understands this notion in terms of the further notion of cognitive virtue, but that it also turns out that one cannot offer a definition of cognitive virtue which does not ultimately itself appeal to the notion of knowledge. Would such a theory of knowledge thereby be a failure? I don’t see why. For we would surely have learnt something important about the nature of knowledge by recognising the truth of this theory, even if ultimately we were not presented with a fully reductive analysis. In particular, we would get to understand the close connections between knowledge and virtue.

6. TWO SOURCES OF SCEPTICISM ABOUT
CONTEMPORARY THEORY OF KNOWLEDGE

This last point offers us some purchase on one particular problem that has been raised for contemporary epistemology. According to this line of objectionmost associated with the work of Timothy Williamson (2000, ch. 1), but not exclusive to himthe theory of knowledge as it is understood by contemporary epistemologists is flawed because it is committed to offering a completely reductive analysis of knowledge. One common reason offered for why this project is problematic is basically a form of pessimistic induction, in that since a wide range of putatively reductive theories of knowledge have been offered since the introduction of the Gettier-style cases to epistemology in the mid-1960s, and since none of them has been universally accepted, so we have reason to think that the very project of decomposing knowledge into its constituent conceptual parts is doomed. A related thought on this score is that the kinds of theories of knowledge being proposed are increasingly complex and ad hoc such that we have grounds to think that this is a degenerating research programme.

Even if we grant that contemporary epistemology is indeed focussed on offering a completely reductive analysis of knowledge, I think we should be very suspicious of this critique. For while there have been some very complex and arguably ad hoc accounts of knowledge in the post-Gettier literature, there have also been some very well-motivated and elegant accounts. In particular, modal accounts of knowledge and virtue-theoretic accounts of knowledge have been very successful at dealing with a range of problem cases and are still ‘live’ proposals in the literature. In both cases they are also usually very straightforward accounts of knowledge, and so can hardly be charged with being unnecessarily complex. Indeed, given the convergence of opinion in epistemology towards views of this sort, one actually has grounds for supposing that if there is a research programme in epistemology which is concerned with offering a completely reductive analysis of knowledge, then it is in a progressive rather than degenerating state.

In any case, as we have seen above, the epistemological project of offering an account of knowledge is not wedded to providing a fully reductive analysis anyway. Indeed, as we also noted, if, say, a virtue-theoretic proposal turned out to be the optimal account of knowledge available, but it also turned out that one could only define cognitive virtues by appeal to knowledge, this needn’t be a catastrophe as the proposal could still be informative. The epistemological project of offering an account of knowledge thus does not stand or fall in terms of whether it can give a fully reductive account of this notion.

A more serious form of scepticism about the contemporary epistemological project comes from a radical form of experimental philosophy, what has been called the negative programme in experimental philosophy. As we have noted above, experimental philosophy in itself poses no fundamental challenge to contemporary epistemology since it merely offers us a way of gaining better empirical data of a kind that is already relevant to the contemporary epistemological enterprise of offering a theory of knowledge. Viewed this way, experimental philosophy offers us a way of enhancing the methodology of contemporary epistemology. Some have argued, however, that the experimental philosophy programme poses a far more serious challenge, not just to contemporary epistemology but also to the whole methodology of analytical philosophy.

In particular, the thought is that the entire appeal to intuitions that is at the heart of analytical philosophy is undermined by the data that has been collected by experimental philosophers. The kind of data which experimental philosophers have in mind here includes, for example, experiments which appear to show that subjects’ responses to cases are highly dependent on supposedly irrelevant factors, such as their ethnicity or the order in which cases are presented to them. The alleged upshot of this data is that analytical philosophers are wrong to rely so much on intuitions, and hence should radically restrict their use of them.

We should notice straight off a flaw in this critique, which is that, as we have argued above, there is much more to the project of analytical epistemology than the appeal to intuitions about cases (i.e., extensional intuitions). Thus even if we grant the import of these experiments, it still doesn’t follow that analytical epistemology is thereby posed anything like the dramaticindeed, potentially fatalthreat envisaged by proponents of the negative programme.

A second issue here is that, as also noted above, analytical epistemology doesn’t put that much epistemic weight on subjects’ first intuitive responses to cases anyway, at least in comparison to their intuitive judgements which are the result of due consideration and reflection. As Williamson puts the point:

“Much of the evidence for cross-cultural variation in judgements on theory experiments concerns verdicts by people without philosophical training. Yet philosophy students have to learn how to apply general concepts to specific examples with careful attention to the relevant subtleties, just as law students have to learn how to analyze hypothetical cases. Levels of disagreement over thought experiments seem to be significantly lower among fully trained philosophers than among novices.” (Williamson 2007, 191)

I think Williamson is entirely right on this score, and I would suggest that this phenomenon reflects the way in which philosophical training provides subjects with the skills relevant to making appropriate intuitive judgements about cases.

Two points need to be noted here. The first is to reiterate the point I made earlier that in responding to the negative programme in experimental philosophy in this way one is not thereby discounting the epistemic weight of subjects’ initial intuitions altogether. For one thing, a substantial divergence between subjects’ initial intuitions and their intuitions under reflection is something which requires explanation.

The second point is that this response poses in effect a further challenge to the negative programmeviz., to run further experiments which challenge even the intuitions under reflection formed by the philosophically trained. We clearly cannot rule-out a priori that a challenge of this sort could be effectively mounted. But for now at least, there is a clear lacuna in the data supporting the negative programme in experimental philosophy.

7. CONCLUDING REMARKS

We have seen, then, that the contemporary epistemological project of offering a theory of knowledge is not nearly as exposed to critique as some have supposed, at least once it is properly described. Of course, there are other challenges to contemporary analytical epistemology besides the two considered here, and I make no claim that even the rendering of the epistemological project that I offer can deal with all the main problems that have been levelled against this project. Still, I hope to have least shown here that there is far more to the methodology of contemporary analytical epistemology than many suppose.

REFERENCES

Alexander, J., & Weinberg, J. (2007). ‘Analytic Epistemology and Experimental Philosophy’, Philosophy Compass 2, 56-80.

Austin, J. L. (1961). ‘The Meaning of a Word’, in his Philosophical Papers, (eds.) J. O. Urmson & G. J. Warnock, 55-75, Oxford: Oxford University Press.

Bartsch, K. & Wellman, H. (1995). Children Talk About the Mind, Oxford: Oxford University
Press.

Bealer, G. (1987). ‘The Philosophical Limits of Scientific Essentialism’, Philosophical Perspectives 1: Metaphysics, (ed.) J. E. Tomberlin, 289-365, Atascadero, CA: Ridgeview.

–(1996a). ‘A Priori Knowledge and the Scope of Philosophy’, Philosophical Studies 81, 121-42.

–(1996b). ‘On the Possibility of Philosophical Knowledge’, Philosophical Perspectives 10: Metaphysics, (ed.) J. E. Tomberlin, 1-34, Atascadero, CA: Ridgeview.

–(1998). ‘Intuition and the Autonomy of Philosophy’, Rethinking Intuition: The Psychology of Intuition and its Role in Philosophical Inquiry, (eds.) M. DePaul & W. Ramsey, 201-40, Lanham, MD: Rowman & Littlefield.

Becker, K. (2007). Epistemology Modalized, London: Routledge.

Bonjour, L. (1998). In Defence of Pure Reason, Cambridge: Cambridge University Press.

Cassam, Q. (2009). ‘Can the Concept of Knowledge be Analysed?’, Williamson on Knowledge, (eds.) P. Greenough & D. H. Pritchard, 12-30, Oxford: Oxford University Press.

Cath, Y. (2011). ‘Metaphilosophy’, Oxford Bibliographies Online: Philosophy,.

Craig, E. (1990). Knowledge and the State of Nature: An Essay in Conceptual Synthesis, Oxford: Clarendon Press.

Daniels, N. (2003). ‘Reflective Equilibrium’, Stanford Encyclopædia of Philosophy, (ed.) E. Zalta, http://plato.stanford.edu/entries/reflective-equilibrium/.

Davidson, D. (1987). ‘Knowing One’s Own Mind’, Proceedings and Addresses of the American Philosophical Association 60, 441-58.

DePaul, M. (1998). ‘Why Bother With Reflective Equilibrium?’, Rethinking Intuition: The Psychology of Intuition and its Role in Philosophical Inquiry, (eds.) M. DePaul & W. Ramsey, 293-309, Lanham, MD: Rowman & Littlefield.

–(2009). ‘Ugly Analyses and Value’, Epistemic Value (eds.) A. Haddock, A. Millar & D. H. Pritchard, 112-38, Oxford: Oxford University Press.
Fricker, M. (2007). Epistemic Injustice: Power and the Ethics of Knowing, Oxford: Oxford University Press.

–(2010). ‘Scepticism and the Genealogy of Knowledge: Situating Epistemology in Time’, Social Epistemology, (eds.) A. Haddock, A. Millar & D. H. Pritchard, 51-68, Oxford: Oxford University Press.

Gelfert, A. (2011). ‘Genealogies of the Concept of Knowledge and Their Cognitive-Ecological Dimension’, Episteme 8, 67-82.

Gendler, T. S. (2007). ‘Philosophical Thought Experiments, Intuitions, and Cognitive Equilibrium’, Midwest Studies In Philosophy 31, 68–89.

Gerken, M. (2009). ‘If the Word ‘Knowledge’ Did Not Exist, Would it Be Necessary to Invent It?’, manuscript.

Gettier, E. (1963). ‘Is Justified True Belief Knowledge?’, Analysis 23, 121-3.

Gilovich, T., Griffin, D., & Kahneman, D. (eds.) (2002). Heuristics and Biases: The Psychology of Intuitive Judgment, Cambridge: Cambridge University Press.

Goldman, A. (2007). ‘Philosophical Intuitions: Their Target, Their Source, and their Epistemic Status’, Grazer Philosophische Studien 74, 1-26.

Goldman, A., & Pust, J. (1998). ‘Philosophical Theory and Intuitional Evidence’, Rethinking Intuition: The Psychology of Intuition and its Role in Philosophical Inquiry, (eds.) M. DePaul & W. Ramsey, 179-200, Lanham, MD: Rowman & Littlefield.

Greco, J. (1999). ‘Agent Reliabilism’, Philosophical Perspectives 13, 273-96.

–(2000). Putting Skeptics in Their Place: The Nature of Skeptical Arguments and Their Role in Philosophical Inquiry, Cambridge: Cambridge University Press.

–(2003). ‘Knowledge as Credit for True Belief’, Intellectual Virtue: Perspectives from Ethics and Epistemology, (eds.) M. DePaul & L. Zagzebski, 111-34, Oxford: Oxford University Press.

–(2007). ‘The Nature of Ability and the Purpose of Knowledge’, Philosophical Issues 17, 57-69.

–(2008). ‘What’s Wrong With Contextualism?’, Philosophical Quarterly 58, 416-36.

–(2009). Achieving Knowledge, Cambridge: Cambridge University Press.

Hazlett, A. (2010). ‘The Myth of Factive Verbs’, Philosophy and Phenomenological Research 80, 497-522.

Henderson, D. (2011). ‘Gate-Keeping Contextualism’, Episteme 8, 83-98.

Hintikka, J. (1999). ‘The Emperor’s New Intuitions’, Journal of Philosophy 96, 127-47.

Ichikawa, J., & Sosa, E. (2010). ‘Intuitions’, Oxford Bibliographies Online: Philosophy, www.oxfordbibliographiesonline.com/display/id/obo-9780195396577-0059.
Jackson, F. (1998). From Metaphysics to Ethics: A Defence of Conceptual Analysis, Oxford: Oxford University Press.

Kahneman, D., Slovic, P., & Tversky, A. (eds.) (1982). Judgment under Uncertainty: Heuristics and Biases, Cambridge: Cambridge University Press.

Kahneman, D., & Tversky, A. (1996). ‘The Reality of Cognitive Illusions’, Psychological
Review 103, 582-91.

Kappel, K. (2010). ‘On Saying that Someone Knows: Themes from Craig’, Social Epistemology, (eds.) A. Haddock, A. Millar & D. H. Pritchard, 69-88, Oxford: Oxford University Press.

Kaupinnen, A. (2007). ‘The Rise and Fall of Experimental Philosophy’, Philosophical Explorations 10, 85-118.

Knobe, J., & Nichols, S. (eds.) (2008). Experimental Philosophy, Oxford: Oxford University Press.

Kornblith, H. (1998). ‘The Role of Intuition in Philosophical Inquiry: An Account with No Unnatural Ingredients’, Rethinking Intuition: The Psychology of Intuition and its Role in Philosophical Inquiry, (eds.) M. DePaul & W. Ramsey, 129-42, Lanham, MD: Rowman & Littlefield.

–(2007). ‘Naturalism and Intuitions’, Grazer Philosophische Studien 74, 27-49.

–(2011). ‘Why Should We Care About the Concept of Knowledge?’, Episteme 8, 38-52.

Kelp, C. (2011). ‘What’s the Point of ‘Knowledge’ Anyway?’, Episteme 8, 53-66

Kusch, M. (2009). ‘Testimony and the Value of Knowledge’, Epistemic Value, (eds.) A. Haddock, A. Millar & D. H. Pritchard, 60-94, Oxford: Oxford University Press.

–(2011). ‘Knowledge and Certainties in the Epistemic State of Nature’, Episteme 8, 6-23.

Lane, M. (1999). ‘States of Nature, Epistemic and Political’, Proceedings of the Aristotelian Society 99, 211-24.

Liao, S. M. (2008). ‘A Defense of Intuition’, Philosophical Studies 140, 247-62.

Lycan, W. G. (2010). ‘Epistemology and the Role of Intuitions’, The Routledge Companion to Epistemology, (eds.) S. Bernecker & D. H. Pritchard, 813-22, London: Routledge.

McDowell, J. (1995). ‘Knowledge and the Internal’, Philosophy and Phenomenological Research 55, 877-93.

–(2009). ‘Wittgenstein’s ‘Quietism’’, Common Knowledge 15, 365-72.

Nagel, J. (2007). ‘Epistemic Intuitions’, Philosophy Compass 2, 792–819.

Neta, R. (2006). ‘Epistemology Factualized: New Contractarian Foundations for Epistemology’, Synthese 150, 247-80.

Nozick, R. (1974). Anarchy, State, and Utopia, New York: Basic Books.

–(1981). Philosophical Explanations, Oxford: Oxford University Press.

Pritchard, D. H. (2005). Epistemic Luck, Oxford: Oxford University Press.

–(2007a). ‘Anti-Luck Epistemology’, Synthese 158, 277-97.

–(2007b). ‘Recent Work on Epistemic Value’, American Philosophical Quarterly 44, 85-110.

–(2007c). ‘The Value of Knowledge’, Stanford Encyclopædia of Philosophy, (ed.) E. Zalta, http://plato.stanford.edu/entries/knowledge-value/.

–(2010). ‘Anti-Luck Virtue Epistemology’, manuscript.

–(2011). ‘The Genealogy of the Concept of Knowledge and Anti-Luck Virtue Epistemology’, Der Begriff des Wissens/The Concept of Knowledge, (ed.) S. Tolksdorf, Berlin: de Gruyter.

–(Forthcoming). ‘Sceptical Intuitions’, Intuitions, (eds.) T. Booth & D. Rowbottom, Oxford: Oxford University Press.

Pritchard, D. H., Millar, A., & Haddock, A. (2010). The Nature and Value of Knowledge: Three Investigations, Oxford: Oxford University Press.

Putnam, H. (1981). Reason, Truth and History, Cambridge, UK: Cambridge University Press.

Rawls, J. (2001). Justice as Fairness: A Restatement, Cambridge, MA: Harvard University Press.

Rorty, R. (1979). Philosophy and the Mirror of Nature, Princeton, NJ: Princeton University Press.

Roush, S. (2005). Tracking Truth: Knowledge, Evidence and Science, Oxford: Oxford University Press.

Rysiew, P. (2008). ‘Rationality Disputes: Psychology and Epistemology’, Philosophy
Compass 3, 1-24.

Shapin, S. (1994). A Social History of Truth, Chicago, IL: University of Chicago Press.

Shatz, M., Wellman, H. M., & Silber, S. (1983). ‘The Acquisition of Mental Verbs: A Systematic Investigation of the First Reference to Mental State’, Cognition 14, 301-14.

Smith, M. (1994). The Moral Problem, Oxford: Blackwell.

Sosa, E. (1991). Knowledge in Perspective: Selected Essays in Epistemology, Cambridge: Cambridge University Press.

–(1998). ‘Minimal Intuition’, Rethinking Intuition: The Psychology of Intuition and its Role in Philosophical Inquiry, (eds.) M. DePaul & W. Ramsey, 257-70, Lanham, MD: Rowman & Littlefield.

–(2007a). A Virtue Epistemology: Apt Belief and Reflective Knowledge, Oxford: Clarendon Press.

–(2007b). ‘Experimental Philosophy and Philosophical Intuition’, Philosophical Studies 132, 99-107.

–(2007c). ‘Intuitions: Their Nature and Epistemic Efficacy’, Grazer Philosophische Studien 74, 51-67.

–(2008). ‘Experimental Philosophy and Philosophical Intuition’, Experimental Philosophy, (eds.) J. Knobe & S. Nichols, 231-40, Oxford: Oxford University Press.

–(2009a). ‘A Defence of the Use of Intuitions in Philosophy’, Stich and His Critics, (eds.) M. Bishop & D. Murphy, 101-12, Oxford: Blackwell.

–(2009b). Reflective Knowledge: Apt Belief and Reflective Knowledge, Oxford: Clarendon Press.

Stich, S. (1988). ‘Reflective Equilibrium, Analytic Epistemology and the Problem of Cognitive Diversity’, Synthese 74, 391-413.

Stich, S., Nichols, S., & Weinberg, J. (2003). ‘Metaskepticism: Meditations in Ethno-Epistemology’, The Skeptics, (ed.) S. Luper, 227-47, Aldershot: Ashgate.

Strawson, P. F. (1992). Analysis and Metaphysics: An Introduction to Philosophy, Oxford: Oxford University Press.

Stroud, B. (1989). ‘Understanding Human Knowledge in General’, Knowledge and Skepticism, (eds.) M. Clay & K. Lehrer, 31-50, Boulder, CO: Westview Press.

–(1994). ‘Scepticism, ‘Externalism’, and the Goal of Epistemology’, Proceedings of the Aristotelian Society (suppl. vol.) 68, 291-307.

Swain, A., & Weinberg, J. (2008). ‘The Instability of Philosophical Intuitions: Running Hot and Cold on Truetemp’, Philosophy and Phenomenological Research 76, 1-20.

Tversky, A., & Kahneman, D. (1974). ‘Judgment Under Uncertainty: Heuristics and Biases’,
Science 185, 1124-31.

Weatherson, B. (2003). ‘What Good are Counterexamples?’, Philosophical Studies 115, 1-31.

Weinberg, J. (2010). ‘Experimental Epistemology’, The Routledge Companion to Epistemology, (eds.) S. Bernecker & D. H. Pritchard, 823-35, London: Routledge.

Weinberg, J., Nichols, S., & Stich, S. (2001). ‘Normativity and Epistemic Intuitions’, Philosophical Topics 29, 429-60.

Wierzbicka, A. (2006). Semantics: Primes and Universals, Oxford: Oxford University Press.

Williams, B. (2002). Truth and Truthfulness: An Essay in Genealogy, Princeton, NJ: Princeton University Press.

Williams, M. (1995). Unnatural Doubts: Epistemological Realism and the Basis of Skepticism,
Princeton, NJ: Princeton University Press.

Williamson, T. (2000). Knowledge and its Limits, Oxford: Oxford University Press.

–(2004). ‘Philosophical ‘Intuitions’ and Scepticism about Judgement’, Dialectica 58, 109-53.

–(2007). The Philosophy of Philosophy, Oxford: Blackwell.

–(2009). ‘Replies to Critics’, Williamson on Knowledge, (eds.) P. Greenough & D. H. Pritchard, 279-84, Oxford: Oxford University Press.

Wittgenstein, L. (1953). Philosophical Investigations, (eds.) G. E. M. Anscombe & R. Rhees, (tr.) G. E. M. Anscombe, Oxford: Blackwell.

Wright, C. J. G. (1992). Truth and Objectivity, Cambridge, MA: Harvard University Press.

Zagzebski, L. (1996). Virtues of the Mind: An Inquiry into the Nature of Virtue and the Ethical Foundations of Knowledge, Cambridge: Cambridge University Press.

–(1999). ‘What is Knowledge?’, The Blackwell Guide to Epistemology, (eds.) J. Greco & E. Sosa, 92-116, Oxford: Blackwell.

NOTES