Social psychologists are mostly "situationists": they claim that a lot of what people do is best explained not by traits of character, but by systematic human tendencies to respond to features of their situations that nobody previously thought to be crucial at all. They think that someone who is, say, reliably honest in one kind of situation will often be reliably dishonest in another. They'd be unsurprised, for example, that Oskar Schindler was mercenary, arrogant, hypocritical, calculating, and vain sometimes … but not always; and that his courage and compassion could be elicited in some contexts but not in others.

Now, to ascribe a virtue to someone is, among other things, to say that she tends to do what the virtue requires in contexts where it is appropriate. An honest person will resist the temptations to dishonesty posed by situations where, say, a lie will bring advantage, or failing to return a lost wallet will allow one to buy something one needs. Indeed, our natural inclination, faced with someone who does something helpful or kind - or, for that matter, something hostile or thoughtless - is to suppose that these acts flow from their character, where character is understood as a trait that is consistent across situations and, therefore, insensitive to differences in the agent's environment, especially small ones. But situationists cite experiments suggesting that small - and morally irrelevant - changes in the situation will lead a person who acted honestly in one context to do what is dishonest in another.

In the past thirty years or so, psychological evidence for situationism has been accumulating. Many of these effects are extremely powerful: huge differences in behavior flow from differences in circumstances that seem of little or no normative consequence. Putting the dime in the slot in that shopping mall raised the proportion of those who helped pick up the papers from one out of twenty-five to six out of seven - that is, from almost no one to almost everyone. Seminarians in a hurry are six times less likely to stop like a Good Samaritan. Mindful of these examples, you should surely be a little less confident that "she's helpful" is a good explanation next time someone stops to assist you in picking up your papers, especially if you're outside a bakery.

Suppose I give you change because (in part) I just got a whiff of my favorite pastry. Of course, if I had a settled policy of never giving change, even that pleasant aroma wouldn't help. So there are other things about me - the sorts of things we would normally assess morally - that are relevant to what I have done. But let's suppose that, other things being equal, if I hadn't had the whiff, I'd have ignored your plaintive plea to stop and change your dollar for the parking meter. Pleased by the ambient aroma, I was inclined to do what, according to the virtue theorist, a kind or helpful or thoughtful person - a virtuous person would do; and I acted on that inclination. A typical virtue theorist will think I have done the right thing because it is the kind thing (and there are no countervailing moral demands on me). But, on the situationist account, I don't act out of the virtue of kindness. Does this act accrue to my ethical credit? Do I deserve praise in this circumstance or not? Have I or haven't I made my life better by doing a good thing?

A situationist might encourage us to praise someone who does what is right or good - what a virtuous person would do - whether or not they did it out of a virtuous disposition, but only for instrumental reasons. After all, psychological theory also suggests that praise, which is a form of reward, is likely to reinforce the behavior. (What behavior? Presumably not helpfulness, but being helpful when you're in a good mood.) But the virtue ethicist cannot be content that one acts as if virtue ethics is true. And we can all agree that the more evidence there is that a person's conduct is responsive to a morally irrelevant feature of the situation, the less praiseworthy it is. If these psychological claims are right, very often when we credit people with compassion, as a character trait, we're wrong: they're just in a good mood. And if hardly anyone is virtuous in the way that virtue ethics conceives of it, isn't the doctrine's appeal eroded? Presumably the presence or absence of the smell of baking is just one among thousands of contextual factors that will have their way with us. There are some philosophers, among them John Doris, who take the social-science literature about character and conduct to pose a serious and perhaps lethal challenge to the virtue ethicist's worldview. If experimental psychology shows that people cannot have the sorts of character traits that the virtue theorist has identified as required for eudaimonia, there are only two possibilities : we have identified the wrong character traits, or we cannot have worthwhile lives. What is the point of doing what a virtuous person would do if I can't be virtuous? Once more, whether I can be virtuous is obviously an empirical question. Once more, then, psychology seems clearly apropos.

Still, we should not overstate the threat that situationism poses. Acquiring virtue, Aristotle already knew, is hard; it is something that takes many years and most people don't make it. These experiments might confirm the suspicion that compassionate men and women are rare, in part because becoming compassionate is difficult. But difficult is not the same as impossible; and perhaps we can ascend the gradient of these virtues only through aspiring to the full-fledged ideal. We could easily imagine a person who, on the virtue ethicists' view, was in some measure compassionate, and who actually welcomed the psychologists' research. Reading about these experiments will only remind her that she will often be tempted to avoid doing what she ought to do. So these results may help her realize the virtue of compassion. Each time she sees someone who needs help when she's hurrying to a meeting, she'll remember those Princeton seminarians and tell herself that, after all, she's not in that much of a hurry ; that the others can wait. The research, for her, provides a sort of perceptual correction akin to the legend you see burned onto your car's rearview mirror: objects may be closer than they appear. Thanks for the tip, she says. To think that these psychological claims by themselves undermine the normative idea that compassion is a virtue is just a mistake. We might also notice what the situationist research doesn't show. It doesn't tell us anything about those 10 percent who were helpful even when rushing to an appointment; perhaps that subpopulation really did have a stable tendency to be helpful or, for all we know, to be heedless of the time and careless about appointments. (Nor can we yet say how the seminarians would have compared with, say, members of the local Ayn Rand society. Virtue ethics is hardly alone in assigning a role to elusive ideals. Our models of rationality are also shot through with such norms. Recall the nineteenth-century hope that, in the formula, logic might be reduced to a "physics of thought." It tells us not how we do reason, but how we ought to reason. And it points toward one way of responding to the question we have posed to the virtue ethicist: how might we human beings take seriously an ideal that human beings must fall so far short of?

You might think the answer is to treat claims about virtues as moral heuristics. But there are many difficulties, I think, for this view. Here is one: for faithful Aristotelians, this whole approach, in which we seek moral heuristics that will guide us imperfect creatures to do what a virtuous person would do, is bound to look very peculiar. Virtue ethics wants us to aim at becoming a good person, not just at maximizing the chance that we will do what a good person would do. Since we're not ideally virtuous, the heuristics model now introduces means-end rationality to maximize your chance of doing what's right by the first test. The trouble is, of course, that virtue ethics requires that we aim at the good for reasons that aren't reducible to means-end rationality. With the cognitive heuristic, what matters is the outcome. But if virtue ethics tells you that outcomes aren't the only thing that matters, then you cannot assess heuristics by means-end rationality - that is, by looking at the probability that they will produce certain outcomes. The value of the virtues does not come just from the good results of virtuous acts or from the enjoyment that virtue produces ; it is intrinsic, not instrumental. A virtuous life is good because of what a virtuous person is, not just because of what she does. It would be a mistake to deny the instrumental significance of honesty; but doesn't our moral common sense recoil at the idea that honesty matters only because of this instrumental significance?

— A. Appiah, Experimental moral psychology (2009)