Character and Consistency

Character and personality traits are invoked to explain what people do and how they live: Peter didn't mingle at the party because he's shy, and Sandra succeeds in her work because she's diligent. Traits also figure in prediction: Peggy will join in because she's impulsive, and Brian will forget our meeting because he's absentminded. So too for those rarefied traits called virtues: James stood his ground because he's brave, and Katherine will not overindulge because she's temperate. Such talk would not much surprise Aristotle (1984: 1l06aI4-23); for him, a virtue is a state of character that makes its possessors behave in ethically appropriate ways, I'll now begin arguing that predictive and explanatory appeals to traits, however familiar, are very often empirically inadequate: They are confounded by the extraordinary situational sensitivity observed in human behavior. Discussion of the descriptive psychology occupies me for several chapters; afterward, I'll be positioned to address related normative concerns.

As I understand it, to attribute a character or personality trait is to say, among other things, that someone is disposed to behave a certain way in certain eliciting conditions. In philosophy, this seems a standard interpretation: Character traits in particular, are widely held to involve dispositions to behavior. So understood, trait attribution is associated with a

conditional:: If a person possesses a trait, that person will exhibit trait-relevant behavior in trait-relevant eliciting conditions. This conditional reflects some natural locutions-- "I thought Andrew was loyal, but if he really was, he would have taken my side at the meeting"--but it inhabits philosophically shaky ground.

Metaphysicians have devoted considerable ingenuity to examples where a dispositional attribution looks true while the associated conditional looks false; some take these examples to show that the conditional approach cannot serve as an analysis of the notion of a disposition. Consider "masking" problems, where a disposition is present together with a countervailing disposition manifest in identical circumstances, that prevents the first disposition from being manifested. Imagine that my crippling shyness prevents my friendliness from being expressed and assume, as seems plausible, the same class of eliciting conditions is relevant to each trait; I have the disposition to friendliness, but the conditional will be false, because my shyness, always trumps my friendliness. This is not simply metaphysical delicacy; people do say things like "he's really a nice person, he's just a little shy" by way of excusing the socially uneasy.

Such examples seem to show that the conditional analysis fails. No bother: I'm doing moral psychology, not metaphysics; my interest is not in conceptual analysis but in the evidential standards governing trait attribution. In outline, the relevant standards are not far to seek. If observed behavior conforms to expectations expressed in the conditional, attribution looks to be warranted; if the expectations expressed in the conditional are unmet, there is pressure to withhold attribution. Think again of the masking problem: Are attributions of friendliness warranted when friendly behavior is invariably blocked by shyness? Perhaps such questions are not easily settled, but it's pretty clear who needs to be doing most of the talking: The burden of proof lies with someone attributing friendliness in the face of repeated, failures to act friendly, while someone asserting the opposite view occupies an enviable rhetorical position. There are certainly metaphysical obscurities plaguing the notion of a disposition, but the conditional neatly articulates a thought that seems plain enough: Attribution of character and personality traits is associated with behavioral expectations.

Talk of traits as dispositions risks vacuity, if it provides explanations no more enlightening than "he acted in this manner because he has dispositions to behave in this manner." In particular, describing virtues as behavioral dispositions is only a very partial accounting; virtue is standardly thought to involve not only what occurs "on the outside" in the form of motives, emotions, and cognitions (see Hardie 1980: 107; Taylor 1988:233). For example, according to McDowell's (1978: 21-3; 1979: 332-3) influential account, virtue is

characterized by a perceptual capacity or reliable sensitivity to ethically salient features of one's surroundings. Virtues are not mere dispositions but intelligent dispositions, characterized by distinctive patterns of emotional response, deliberation, and decision as well as by more overt behavior.

This is not to say that behavior is inconsequential: "His ethical perceptions were unfailingly admirable, although he behaved only averagely" is an uninspiring epitaph. In Fitzgerald's Gatsby, Carraway puts it pointedly: "Conduct may be founded on the hard rock or the wet marshes, but after a certain point I don't care what it's founded on." Carraway's exasperation seems a bit willful, but he's right to say that there's no getting around the question of behavior, whatever psychological story gets told. I doubt any writer on virtue is seriously inclined to deny this. As Aristotle (1984: 1098b30-1099a5) observes, the activity, not the possession, of virtue is paramount; possession without activity means a life where nothing virtuous gets done. Accounts of virtue emphasizing "internal" processes are best understood as complements, not competitors, to characterizations emphasizing behavioral dispositions, inasmuch as they explicate the psychological processes subserving behavior. For example, McDowell (1979: 332-3, 343-6) maintains that virtue's characteristic perceptual capacity produces ethically appropriate conduct; he invokes virtues to explain the virtuous person's behavior. While overt behavior is only one facet of interest for a moral psychology of traits, it is certainly of central interest.

I should say something more about the behaviors or, rather, patterns of behavior at issue. According to Aristotle (1984: 1105a27-b1), genuinely virtuous action proceeds from "firm and unchangeable character" rather than from transient motives. The virtues are hexeis (1984: 1l06all-12), and a hexis is a state that is "permanent and hard to change" (1984: Categories, 8b25-9a9). Accordingly, while the good person may suffer misfortune that impairs his activities and diminishes happiness, he "will never [oudepote] do the acts that are hateful and mean" (1984: 1l00b32-4; cf. 1128b29; d. Cooper 1999: 299nI4). The presence of virtue is supposed to provide assurance as to what will get done as well as what won't; for Aristotle (1984: 110 1 a 1-8; 114oa2 6-b30) , the paradigmatic ally virtuous Phronimos, or practically wise man, is characterized by his ability to choose the course of action appropriate to whatever circumstance he is in, whether it be easy or excruciating. Arete, Aristotle says, is "always concerned with what is harder" (1984: 1105a8-10); standing firm in the most terrifying crises, not just in any frightening situation, is diagnostic of the brave person (1984: 1115a24-6; 1117a24'-22).

These features of Aristotle's moral psychology are prominent in contemporary' virtue ethics. McDowell (1978: 26-7) contends that considerations favoring behavior contrary to virtue are "silenced" in the virtuous person; although she may experience inducements to vice, she will not count them as reasons for action. Again, good character provides positive as well as

negative assurance; according to Dent (1975: 328), virtue effects appropriate behavior in "ever-various and novel situations," while for McDowell (1979: 331-3), genuine virtue is expected to "produce nothing but right conduct." As I put it, virtues are supposed to be robust traits; if a person has a robust trait, they can be confidently expected to display trait-relevant behavior across a wide variety of trait-relevant situations, even where some or all of these situations are not optimally conducive to such behavior. I've already burdened the text with too many quotations, but let me emphasize that the above selections are quite representative: An emphasis on robust traits and behavioral consistency is entirely standard in the Aristotelian tradition of character ethics.

Consistency alone does not a virtue make, as can be seen with another notion that has had some ethical currency, integrity. The term is sometimes used as a highly general term of approbation: Saying that a person has integrity can mean something similar to saying that they are a good or admirable person. But integrity can figure in a life that is morally suspect or even morally reprehensible; the Nazi who cannot be bribed to spare Jews very arguably displays integrity. As the etymological origin of "integrity" in the Latin integritas or "wholeness" tempts one to put it, the incorruptible Nazi manifests a unity between his reprehensible principles and his loathsome deeds; rather than being swayed by financial inducements, he consistently acts in accordance with his values. Whatever one cares about, this unity is necessary for executing one's projects in the face of obstacles; without integrity, it is only by luck that one's values will come to pass in their life. A stiff spine alone is not enough for a good life, but it is difficult to imagine a good life without one. Relatedly, to say that someone "has character" is not necessarily to express wholehearted approbation, as attributions of virtue (when offered without irony) typically are, but the locution will, it seems to me, generally carry at least grudging admiration; a bad man who stands up for what he believes can in this respect display estimable character, while falling dismally short of virtue. For my purposes, the thing to note is that the ethical interest of notions like robustness and behavioral consistency extends well beyond their association with virtue.

A preoccupation with. behavioral consistency is not limited to ethics; it is equally evident in personality psychology. According to Pervin (1994a: 108), a personality trait is "a disposition to behave expressing itself in consistent patterns of functioning across a range of situations," while Brody (1988: 31) understands personality traits as "personal dispositions to behave in comparable ways in many diverse situations" (cf. Mischel 1968: 8-9; Wright and Mischel 1987: 1159-61; Ajzen 1988: 34-7). Once again, talk of traits as dispositions invites a conditional: If a person possesses a trait, that person will exhibit trait-relevant behavior in trait-relevant eliciting conditions. This

initial formulation is inadequate; it seems to imply that traits will exceptionlessly issue in trait-relevant behavior, but sporadic failures of trait-relevant behavior probably shouldn't be taken to disconfirm attributions. Then the conditional should be formulated probabilistically: If a person possesses a trait, that person will engage in trait-relevant behaviors in trait-relevant eliciting conditions with probability p. Not just any probability will do; probabilities of slightly above chance do not underwrite confident attributions of a trait. The conditional should be amended once more: If a person possesses a trait, that person will engage in trait-relevant behaviors in trait-relevant eliciting conditions with markedly above chance probability p. Now "markedly above chance" is not a locution of admirable precision. In a general statement, this imprecision is unavoidable, because the degree of predictive confidence associated with a trait attribution may vary according to the trait and individual in question. Fruitful argument must attend to particular instances, but insofar as it is safe to say that a chance probability of trait-relevant behavior is no evidence for attribution, it is quite plausible to think that in the generality of cases, the probability must substantially exceed chance.

Not every consistent behavior pattern is telling evidence for trait attribution: If someone consistently behaves gregariously across a run of situations where most everyone would, their behavior is not decisive evidence for extraversion. Rather, it is individuating behavior--behavior that is outside the population norm for a situation--that counts as evidence for trait attribution. Actually, individuation per se is not the issue; in principle, every individual in a population could possess a trait. Individuation is evidentially significant because where trait-relevant behavior varies markedly in a situation, there is reason to think that the situation is less than optimally conducive to that behavior. Situations of this sort are diagnostic: unfavorable enough to trait-relevant behavior that such behavior seems better explained by reference to individual dispositions than by reference to situational facilitators. Behavioral consistency across a run of situations, where at least some of the situations are diagnostic, is the evidence required for attribution of person�ality traits. Like virtues, personality traits are supposed to be robust.

It's clear that philosophers' talk of character and psychologists' talk of personality exhibit substantial affinities, as psychologists' not infrequent use of "character" in discussions of traits and personality might lead one to suspect. Still, I should take a bit of care over the differences. Character traits appear to have an evaluative dimension that personality traits need not; for example, the honest person presumably behaves as she does because she values forthrightness, while the introvert may not value, and may in fact disvalue, retiring behavior in social situations. This is not to say that acting according to a value requires a conscious belief, and it is still less to say that the virtuous person must act "in the name" of virtue (see Williams 1985: 9-10). But if a person maintains a value, she will be expected to voice at least recognizable variants of its characteristic considerations when asked to rationalize

p.20

her behavior; if a person has a virtue, the relevant evaluative commitments can be expected to surface under "evaluative cross-examination."

The thought that virtues have this sort of evaluative dimension is respectably Aristotelian; Aristotle (1984: 1105a30-b1) maintains that genuinely virtuous activity is undertaken knowingly and for its own sake. It is less clear that this thought neatly applies to vices and other negatively valenced traits of character; is cowardly behavior necessarily tile expression of the actor's values? This is a reasonable concern, but I'll insist that cowardice and other negatively valenced character traits do involve the relevant sort of evaluative dimension; perhaps the coward values safety more than honor, loyalty, and dignity. To get tolerably clear on this would take rather more discussion of evaluation, and the evaluations associated with each trait of character than I'm going to provide. For my interest is not so much what distinguishes character and personality traits as what they have in common: behavioral consistency as a primary criterion of attribution.

My approach may seem simplistic even with regard to behavioral consistency: It appears to be a "one size fits all" account, deaf to differences amongst individual traits, while different traits may have different attributive standards. For example, conceptions of traits like courage and loyalty appear to have high standards for behavioral reliability "built in," while predictive confidence regarding displays of compassion may be considerably lower; loyalty may require unfailing fealty to obligation, while compassion may require only engaging a certain percentage of opportunities for compassionate behavior (see Hunt 1997: 65-6, 89-91 ). Moreover, attribution of negatively valenced traits may require very little in the way of behavioral consistency; perhaps one doesn't have to reliably falter, but only sporadically falter, to be counted a coward.As a rough and ready generalization, I'm inclined to say that "generic" character and personality traits are typically expected to be less robust than virtues; in the exalted realm of virtue, attributive standards may be substantially more demanding. But matters are best put to the test by consideration of particular traits and concrete cases, something I'll need to do a good bit of as we go along. At present, please note that the difficulty I wish to press does not involve violations of some absolute and general standard of consistency for trait attribution; I do not suppose that any such standard exists. Rather, the difficulty is that for important examples of personality and character traits, there is a marked disparity between the extent of behavioral consistency that familiar conceptions of the trait lead one to expect and the extent of behavioral consistency that systematic observation suggests one is justified in expecting.

The Inseparability of the Virtues

Aristotelian moral psychology involves not only a view about the nature of character organization. Aristotle

p.21

(1984: 1144b30-1145a2) maintains a reciprocity thesis, the view that "you have one of the virtues of character if and only if you have them all" (see Irwin 1988: 61), while contemporary writers like McDowell have endorsed a unity thesis, where the apparently discrete virtues turn out to be different manifestations of a "single complex sensitivity" (1979: 333; cf. Murdoch 1970: 57-8). Such claims have struck many commentators as badly contrary to fact (e.g., Flanagan 1991: 4-11). It is easy to imagine a person who is, say, courageous and intemperate; indeed, it. is tempting to think that such a person is courageous in part because she is intemperate.

However, the seemingly implausible inseparability thesis is motivated by two quite plausible claims: first, that virtue reliably secures ethically appropriate conduct and, second, that evaluative considerations are interdependent. (see McDowell 1979: 332) For example, I cannot know whether I should now battle unto death if I do not know whether the cause I champion is a just one; courage untempered by justice may effect conduct that is stupid, brutal, or both. With a little thought, such possibilities multiply: Justice is constrained by compassion, compassion by justice, and similarly for the other virtues; it becomes tempting to think that the full realization of one virtue requires the full realization of them all.

Skepticism lingers - one might reject either of the motivating claims. Perhaps virtue can lead us to do wrongly, as when a good person loyally carries out the orders of a bad one. Or perhaps practical problems are in a sense atomistic; the simple hoplite can be genuinely courageous while knowing nothing of justice. A more modest relative of the inseparability thesis holds only that virtues and vices cannot coexist in a single personality, because genuine instances of virtues manifest evaluative commitments that preclude vices (see Kraut 1988: 83). This exclusionary thesis demands less than the inseparability thesis; while it insists that a courageous person will not be cruel, it allows that he may not be especially compassionate. But it is bold enough; although it may seem trivially true that the presence of courage rules out the corresponding vice of cowardice, why think it rules out an apparently unrelated vice like illiberality?

More sensitive to stubborn separatist intuitions is a "limited" inseparability thesis; this allows for separability of virtues across different domains of practical endeavor but asserts that virtues are inseparable within a given practical domain (Badhwar 1996: 307-8). Recognizing the domain-specificity of practical endeavor helps explain how the upstanding public servant can be a faithless husband; the marital and the political are different practical domains and may engage very different cognitive, motivational, and eval�uative structures. We can also understand how there may be considerable integration within a practical domain; a scholar must be both diligent and honest in her research if she is to do commendable work, although this does not entail that she exhibit the same qualities in her teaching. Domain specificity is important, and it will inform my own view later on, but notice

that as it stands the limited inseparability thesis involves a strong demand for integration: If the public servant is honest in the domain of her work, she will also be compassionate in that domain (see Badhwar 1996: 321-5). Once again, this may strike us as contrary to fact: The Queen intoning "Let them eat cake" isn't prevaricating, but it ain't compassion, either. Indeed, I eventually argue that there is good empirical reason to think that conflicting traits are frequently manifested within limited practical domains.

I'll keep in mind, though, that claims of inseparability make an elusive target for empirical attacks. Defenders may claim that inseparability holds only for perfect virtue; they can thereby allow the abundant appearances of separability and simply insist that these cases involve something less than the full realization of virtue (see Aristotle 1984: 1144b35-114Sa2; Irwin 1997: 193). This expedient apparently removes inseparability from empirical threat; since we can expect perfect virtue to be extremely rare, neither a paucity of cases suggesting inseparability nor a plethora of cases suggesting separability need give defenders of inseparability pause. However, I think that characterological moral psychology is very often rather more empirically ambitious than this suggests. Irwin (1997: 213) remarks that Aristotelian standards of inseparability seem "neither unrealistic nor unreasonable," while Badhwar (1996: 317) understands limited inseparability as "an empirical thesis, subject to revision by developments in psychology." Questions regarding the empirical commitments of characterological moral psychology are delicate and various, and I must postpone fuller discussion until much later in the day. For the moment, it suffices to say that the literature provides some encouragement for empirically evaluating notions of inseparability.

Globalism

The conception of character at issue, which I'll call globalism, can now be stated a bit more precisely. Globalism maintains the following three theses, two regarding the nature of traits and the third regarding personality organization:

(1) Consistency. Character and personality traits are 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.

(2) Stability. Character and personality traits are reliably manifested in trait-relevant behaviors over iterated trials of similar trait-relevant eliciting conditions.

(3) Evaluative integration. In a given character or personality the occur�rence of a trait with a particular evaluative valence is probabilistically related to the occurrence of other traits with similar evaluative valences.

Taken together, these theses construe personality as more or less coherent and integrated with reliable, relatively situation-resistant behavioral

p.23

implications. Or, more pithily: Globalism construes personality as an evaluatively integrated association of robust traits. We are justified in inferring globalist personality structures if behavior reliably exhibits the patterns expected on the postulation of such structures: In the first instance, runs of trait�relevant behavior should exhibit consistency across situations (intratrait consistency); in the second, these runs of consistent behavior will exhibit evaluative affinities with other such runs (intertrait consistency). The honest person, for example, will be consistently honest, and will also exhibit consistent behavior indicative of traits related to honesty, such as loyalty and courage.

As the preceding discussion suggests, both characterological moral psychology and personality psychology are typically committed to the first two theses, consistency and stability. The idea of evaluative integration is rather less prominent in personality psychology than in character ethics, and even in character ethics, the comprehensive integration required by the inseparability and unity theses has been the object of suspicion. Moreover, the theses are detachable: Neither consistency nor stability entails integration, and stability does not entail consistency. It also seems to me that they are differentially plausible; while I reject the consistency and evaluative integration theses, I myself will endorse a variant of stability. Then argument regarding the three globalist theses must to a certain extent proceed independently, but all three theses have been associated with Aristotelian approaches to moral psychology; at least initially, there is good reason for thinking of them together. These qualifications made, I'll state my central contention in descriptive moral psychology. Systematic observation typically fails to reveal the behavioral patterns expected by globalism; globalist conceptions of personality are empirically inadequate. This is not to repudiate every aspect, or all variants, of characterological moral psychology and personality psychology; I mean only to quarrel with commitments to the empirically inadequate aspects of globalism. Of course, since I think that many participants in these endeavors exhibit very substantial globalist commitments, the quarrel is not inconsequential. But there may well be people working in both areas who avoid doing so. To them, I apologize in advance, for I sometimes lapse into locutions like "characterological" and "personological" when I intend only approaches committed to globalism. But again, I think globalism runs far and wide through both characterological moral psychology and personality psychology; with my apology made, I may on occasion omit the qualification.

Situationism

While globalist conceptions of personality have not received much scrutiny in ethical theory, they have long been the subject of rancorous debate in social and personality psychology. The crisis came in 1968 with the

p.24

publication of two critical assessments of personality psychology by Mischel and Peterson. While they expressed similar views, Mischel (1968: 146) garnered the lion's share of discussion--and criticism--with the frank assertion that globalist conceptions of personality traits are "untenable." Mischel's basic argument is simple: Globalist conceptions of personality are predicated on the existence of substantial behavioral consistency, but the requisite consistency has not been empirically demonstrated (Mischel 1968: 6-9,146-8). Subsequent controversy notwithstanding, I think Mischel's argument is still a good one, and it should by now be clear that I endorse something like it. But if I'm to come by this endorsement honestly, I'll have to provide a fuller explication.

The story really began long before Mischel with Hartshorne and May's (1928) monumental "Character Education Inquiry," a comprehensive empirical study of honest and deceptive behavior in schoolchildren. Hartshorne and May (1928: I, 385) found that even across quite similar situations, honest and dishonest behavior were displayed inconsistently; they concluded that honesty is not an "inner entity" but is instead "a function of the situation." Shortly thereafter, their contention was buttressed by Newcomb's (1929) study of introversion and extraversion in "problem boys," which found that trait-relevant behaviors were not organized into con�sistent patterns but instead were highly situation-specific and inconsistent. The difficulty and expense of extensive behavioral observation has generally prohibited exhaustive study in the vein of Hartshorne and May and Newcomb (see Ross and Nisbett 1991: 98), but where relevant behavioral research has been conducted, degrees of situational sensitivity in behavior that confound globalist constructions of personality are typical. More than twenty years after Mischel's 1968 study, Ross and Nisbett's (1991: 2-3, 97) review echoed his conclusion: Existing empirical evidence for globalist conceptions of traits is seriously deficient.

It is this research tradition that has come to be known as "situationism." Situationism's three central theoretical commitments, amounting to a qualified rejection of globalism, concern behavioral variation, the nature of traits, and personality organization.

(1) Behavioral variation across a population owes more to situational differences than dispositional differences among persons. Individual dispositional differences are not so behaviorally individuating as might have been supposed; to a surprising extent it is safest to predict, for a particular situation, that a person will behave in a fashion similar to the population norm (Ross and Nisbett 1991: 113).

(2) Systematic observation problematizes the attribution of robust traits. People will quite typically behave inconsistently with respect to the attributive standards associated with a trait, and whatever behavioral consistency is displayed may be readily disrupted by situational variation. This is not to deny the existence of stability: the situationalist acknowledges that

p.25

individuals may exhibit behavioral regularity over iterated trials of substantially similar situations (Ross and Nisbett 1991: 101; cf Wright and Mischel 1987: 1161-2; Shoda, Mischel, and Wright 1994: 681-3).

(3) Personality is not often evaluatively integrated. For a given person, the dispositions operative in one situation may have an evaluative status very dif�ferent from those manifested in another situation; evaluatively inconsistent dispositions may "cohabitate" in a single personality.

In sum, situationism rejects the first and third globalist theses, consistency and evaluative integration, while allowing a variant of the second, stability.

In my interpretation, situationism does not entail an unqualified skepticism about the personological determinants of behavior; it is not a Skinnerian behaviorism. Although reflection on situationism has caused me to reject an understanding of behavior as ordered by robust traits, I allow for the possibility of temporally stable, situation-particular, "local" traits that are associated with important individual differences in behavior. As I understand things, these local traits are likely to be extremely fine-grained; a person might be repeatedly helpful in iterated trials of the same situation and repeatedly unhelpful in trials of another, surprisingly similar, situation. The difficulty for globalism is that local traits are not likely to effect the patterns of behavior expected on broad trait categories like "introverted," "compassionate," or "honest." Even seemingly inconsequential situational variations may "tap" different dispositions, eventuating in inconsistent behavior. I argue that systematic observation of behavior, rather than suggesting evaluatively integrated personality structures, suggests instead fragmented personality structures--evaluatively disintegrated associations of multiple local traits (see Doris 1996; 1998: 507-8).

Unfortunately, situationism is sometimes the victim of caricature. Situationists have been accused of foul play in setting the standards for trait theory, allegedly "demanding consistency of behavior in every single situation" (Epstein 1990: 96). This could hardly apply to the approach I've taken here, which proposes probabilistic standards for trait attribution. Critics have also charged situationists with making grandiose claims for the power of situations; Funder and Ozer (1983: 111) claim that the situationists often take their experiments to suggest that "correlations between measurable dimensions of situations and single behaviors typically approach 1.0." Neither I nor any situationist I know of maintains such an unlikely view, which is tantamount to saying that individual differences are in general very nearly irrelevant to behavioral outcomes. Indeed, I've just proposed an account of personality that acknowledges individual dispositional differences.

The situationist must acknowledge with Mischel (1968:8) that "previous experience and genetic and constitutional characteristics affect behavior and result in vast individual differences among people," while the personality theorist must acknowledge with Allport (1966:2) that situations have extraordinary effects on behavior. We must take care to avoid needlessly

p.26

polarizing the debate; all parties should agree that behavioral outcomes are inevitably a function of a complex interaction between organism and environment (Bem and Funder 1978: 485-6).

It likewise courts misunderstanding to suppose that situationism is embarrassed by the considerable behavioral regularity that undoubtedly is observed; because the preponderance of people's life circumstances may involve a relatively structured range of situations, behavioral patterns are not, for the most part, radically disordered (see Mischel 1968: 281). Still, there is reason to doubt that behavioral regularity is as substantial as casual observation may suggest. Every person, in the course of his or her life, exhibits a multitude of behaviors; since social observation is usually piecemeal and unsystematic (even intimates may be observed on occasions limited in both number and diversity), observers should be hesitant to take a limited sampling of behaviors as evidence for confident interpretations of personality. At bottom, the question is whether the behavioral regularity we observe is to be primarily explained by reference to robust dispositional structures or situational regularity (see Hannan 1999). I insist that the striking variability of behavior with situational variation favors the latter hypothesis.

Personality, Behavior, and Evidence

One response personality psychologists have made, when comparing their theoretical models with the messy and multifarious quality of human behavior, is to deemphasize the importance of overt behavior in the measurement of personality. Allport argued (1966: 1; cf. 1931) that "[a]cts, and even habits, that are inconsistent with a trait are not proof of the non-existence of the trait." It is certainly right to say, given the importance of psychological dynamics to thinking about personality, that behavior alone is not diagnostically decisive. But if one leans too hard on this observation, one may start to say things that sound rather dubious; if a habit that is contrary to a trait does not undermine the attribution, it is hard to see what possibly could. Personality theory too lenient in its behavioral criteria is, because of its lack of empirically testable hypotheses, "unfalsifiable" (see Popper 1959a; Mischel 1968: 56; Kenrick and Funder 1988: 24).

A central reason for the neglect of overt behavior in personality psychology has been the difficulty and expense of systematic behavioral observation. The standard and substantially cheaper alternative is "paper and pencil" personality assessment based on subjects' self-reports, where investigators have found regularities in test responses favorable to standard theoretical constructions of personality (Ross and Nisbett 1991: 98; Holzman and Kagan 1995: 5-8). Then it is hardly surprising that personality psychologists might pursue self-report measures at the expense of behavioral measures and insist, with Allport, that behavioral measures are not a decisive factor in trait attribution. Unfortunately, there is often "a minimal correlation, or none at

all" between self-report and overt behavioral measures of traits (Holzman and Kagan 1995: 5), so that the competing approaches to personality are likely to generate conflicting conclusions. It sometimes appears as though two discrete disciplines have emerged, one of self-report taxonomies and one of overt behavioral measures, with very different standards of success and, accordingly, very different results (see Epstein 1983: 361).

Judiciously interpreted responses to well-designed paper and pencil instruments can tell us much about what people are like; not all significant differences among persons are neatly related to overt behavior. But psychologists, like the rest of us, are interested in the behavioral implications of personality. Allport (1966: 1; d. 1931), despite the hedge we have just considered, conceives of traits as "determinative" in behavior, while Tellegen ( 1991: 12) insists that "the contribution of traits to behavior makes a difference in life." This difference, I think, is fairly understood as involving more than results on pencil and paper personality inventories. A viable psychology of personality must rely on behavioral observations as well as self-report data; if trait attributions are to help tell us how people will get on in the world, they must be shown to have behavioral implications (see Holzman and Kagan 1995: 9).

Nevertheless, an emphasis on overt behavior should not be allowed to obscure conceptual difficulty, especially regarding which behaviors and situations are relevant to a trait. Only when there are at least rough and ready answers to these questions can empirical investigation be profitably pursued, yet the problem of fully specifying the conditions relevant to manifestation of a disposition is notoriously complex (see Ryle 1949: 116 ff.; Brandt 1970: 29-30). Fortunately, useful empirical work does not require full specification; so long as it is possible to identify conditions of uncontroversial relevance to a trait, empirical investigation of behavior in these conditions can be more or less conceptually untroubled. In the next chapter, I'll proceed in this fashion with the ethical trait I take as my central test case, compassion. I'll identify conditions that are obviously relevant to compassion, and since these conditions have been the subject of empirical study, I'll be able to put globalist conceptions of compassion to empirical test. If I'm right about what they show, such tests are powerful motivation for skepticism about traditional notions of character.