John Doris has recently argued that the conception of character employed in philosophical ethics should be revised. We must abandon the idea that behaviour is regulated by such ‘global’ traits as honesty, generosity, compassion, and courage, and think instead in terms of such ‘local’ traits as ‘sailing-in-rough-weather-with-one's-friends-courage’. The specification of each trait should refer to an array of situational features. This view is becoming influential: Peter Goldie has endorsed it, and Maria Merritt has formulated a variant according to which the situational features relevant to the specification of a trait are always social features.Footnote ¹

We must adopt this ‘fragmentary’ theory of character, its proponents argue, in order to explain the data from certain experiments in social psychology. The subjects in each of these experiments were presented with opportunities for compassionate behaviour, but surprisingly few took these opportunities under such countervailing pressures as the presence of a passive bystander or of a polite experimenter in a technician's coat asking them to deliver seemingly dangerous electric shocks to what appeared to be another volunteer. One kind of experiment shows that the likelihood of a given subject behaving compassionately decreases in inverse proportion to the degree of hurry that subject is in. This data, the argument runs, indicates that behaviour is not determined by such traits as compassion, but by such traits as in-a-hurry-compassion, on-time-compassion, time-to-spare-compassion, presence-of-a-passive-bystander-compassion, defying-authority-compassion, and so on.Footnote ²

This argument has met with two kinds of response. One is that the experiments are not appropriate for drawing these kinds of conclusions.Footnote ³ The other is that mainstream theories of virtue are anyway not committed to the claim that most people's behaviour is regulated by such unified global traits as compassion.Footnote ⁴ Both of these claims are debatable, but this debate would be unnecessary. For what both responses miss is that the data as they stand are perfectly compatible with the idea that the subjects’ behaviour was regulated by global traits so long as we understand these in the right way, and that understanding traits this way is pragmatically preferable to adopting the fragmentary theory.

Explaining the data in terms of global traits is fairly straightforward. While each experimental situation presented an opportunity for compassionate behaviour, each also contained pressures that might elicit contrary behaviour manifesting a different trait, such as punctuality, obedience to authority, or concern for peer approval. There is no reason to think that one situation can invite the manifestation of only one trait, and philosophers have not normally thought this to be the case. Situations are complex and issue conflicting invitations. One's behaviour depends on which trait is strongest and which invitation most forceful.

Proponents of the fragmentary theory fail to see this because they conceive of global traits in purely behaviouristic terms: Doris writes that such traits would be ‘reliably manifested in trait-relevant behavior across a diversity of trait-relevant eliciting conditions that may vary widely in their conduciveness to the manifestation of the trait in question’; Merritt claims that they would ‘reliably give rise to the relevant kind of behavior, across the full range of situations in which the behavior would be appropriate, including situations that exert contrary pressures’; and Goldie writes that ‘honest people can be relied on to act honestly wherever honesty is appropriate’.Footnote ⁵

These claims are far from perspicuous. If the terms ‘trait-relevant’ or ‘appropriate’ are meant to indicate situations inviting the manifestation of that trait and in which there are no countervailing pressures, then the data cited would present no evidence against the idea that the subjects possessed relatively stable levels of compassion, as we have seen. So the terms must mean, as Merritt's wording suggests, situations inviting the manifestation of that trait, regardless of whether there are countervailing pressures. Their argument therefore requires that traits be understood purely in terms of stimulus and response, and not in terms of any inner mediating mental items such as inclinations that may be elicited yet not manifested in action.

We might argue at this stage that the proponents of the fragmentary theory are simply employing the wrong notion of traits. We might argue that classic discussions in the ethics of virtue, such as Aristotle's discussion of strength and weakness of will, or the debates over the unities and conflicts among the virtues, involve the idea that a single situation can appeal to more than one trait. We might even point out that global traits as defined by the proponents of the fragmentary theory are remarkably similar to Aristotle's vices of excess, and that since he considers all virtues and vices to be character traits, he cannot understand traits in that way.

Such a response would not, however, provide us with a positive reason to retain this construal of global traits in terms of inner mediating items. But if we consider the practical consequences of adopting the behaviourist notion and therefore the fragmentary theory of traits, we find such a reason: the behaviourist notion unnecessarily impedes pursuit of the ethical aims for which we need a theory of character. There are broadly two such ethical aims: character-development and situation-management.

The former can be grounded in any of a variety of metaethical theories. One might want to inculcate certain traits on the grounds that the most basic moral judgements are concerned with character. If one holds this position, one might either take the Aristotelian view that these judgements do not depend on any non-moral judgements, or take the opposing view that a judgement of character depends on non-moral judgements about the likely consequences of possessing those traits. Alternatively, one might argue that the most basic moral judgements are concerned with right action, but hold that the possession of certain traits is desirable because it makes right action more likely. This position is neutral between the Kantian view that whether an action is right depends on the motivation behind it, and the utilitarian view that it depends on the action's consequences. Whichever metaethical grounding it has, pursuit of the aim of character-development requires one to think of traits either in the behaviourist or in the non-behaviourist way.

Doris and Merritt present the aim of situation-management as an alternative to character-development.Footnote ⁶ One might, however, hold a hybrid position and claim that a combination of the two approaches will best promote either good consequences or one's doing one's duty. Either way, avoiding morally dangerous circumstances and seeking out situations in which one is likely to behave well involves the ability to predict how one is likely to behave in a given situation, which requires an understanding of one's character, which must itself involve either the behaviourist or the non-behaviourist conception of traits.

If we choose the behaviourist conception of traits, then the reflection involved in the pursuit of either of these ethical aims is extremely complicated, whereas if we retain the non-behaviourist conception it is relatively simple. The earlier list of local traits resulting from the behaviourist interpretation of the experimental data is disingenuous, since each refers only to two situational features where real situations can be far more complicated. Accepting the idea of local traits should not lead to ascribing such traits as in-a-hurry-compassion and defying-authority-compassion, but to ascribing such traits as compassion - defying - authority - in - the - presence - of - a - passive - bystander - who - is - not - a - prior - acquaintance - when - not - in - a - hurry - and - where - the-person-in-need-of-help-is-not-an-acquaintance. Even this may be too simple: the range of situational features to be built into every trait may include many not yet considered. The behaviourist conception leads to a massive proliferation, a combinatorial explosion of possible traits. The ethical aims of character-development and situation-management would therefore require one to maintain a vast database of detailed specifications of all the situations one has faced and of one's behaviour in those situations.

The non-behaviourist conception, on the other hand, allows us to look for patterns in our behaviour, and reflect quite simply that we are overly obedient to authority, or too fastidious in our concern for punctuality, or insufficiently concerned about the impact of our actions on the well-being of others. The ethical aims of character-development and situation-management are therefore far more manageable if we retain this conception than if we replace it with the behaviourist one. Since the two conceptions are equally empirically respectable, it seems that we should opt for the one that best serves our ethical purposes in having a theory of character, and therefore retain the non-behaviourist idea of global traits rather than replace it with the behaviourist conception of traits and therefore the fragmentary theory of local traits.Footnote ⁷