The Person and the Situation

by Lee Ross

What we seek to establish is that laypeople consistently fail to make sufficient allowance for the role that construal plays in determining behavior, a failure with profound personal and social consequences.

The first error is a failure to recognize the degree to which one’s own understanding of stimuli is the result of an active, constructive process, rather than a passive reception and registering of some external reality.

For instance, our suffering is redemptive.

The second error is the failure to appreciate the inherent variability of situational construal. The way any two people interpret a given situation, or even the way a particular person interprets identical stimuli on two different occasions, is only imperfectly predictable and is always uncertain to some degree.

The third error concerns causal attributions for behavior. People fail to recognize the extent to which observed actions and outcomes, especially surprising or atypical ones, may prove to be diagnostic not of the actor’s unique personal dispositions but rather of the objective situational factors facing the actor and of the actor’s subjective construals of those factors.

This is a helpful and lived distinction.

A much more sensible convention for defining effect size was suggested by Cohen (1965, 1977), who suggested that the magnitude of experimental effects should be judged relative to the variability of the measure in question. Thus, by Cohen’s criterion, a difference between two means that corresponded to a quarter of a standard deviation in the distribution of the relevant measure would be deemed small, a difference corresponding to half a standard deviation would be deemed moderate, and a difference corresponding to a whole standard deviation would be deemed large. This statistical definition, and other related ones, assess effect size relative to all nonspecified, “random” determinants of variability or, in other words, relative to “noise.” The definition effectively finesses, in fact ignores, all considerations of the nature of the variable under consideration and the units of measurement involved. Therein lies both its major virtue and, as our discussion of the other two criteria will make clear, also its chief drawbacks.

Similarly, as we will discuss in more detail in Chapter 4, a personality test that was inexpensive to administer and that could predict “only” 10 percent of the variance in some important outcome, could prove to be very valuable and “cost-effective” for many familiar assessment or prediction tasks, for example, for selecting people who are likely to be extreme on some dimension (see Abelson, 1985).

More specifically, we will provide a “situationist” and “subjectivist” account of individual differences – one that gives heavy weight to the complex dynamics of social systems and to the role of construal processes. Our goal will thus be an account of individual differences that seeks to explain what kinds of differences are likely to exist and be important, when they are likely to be obscured, and when misinterpretations of such differences are likely to arise.

More importantly, we will stress the fact that people may characteristically fail to recognize and make allowance for the vagaries of construal, both in predicting their own behavior and in predicting and interpreting the behavior of others.

The how of what we do

The mob situation, they noted, seems at once to energize the individual participants and to rob them of the rationality and sense of propriety that otherwise guide their behavior.

In fact, anywhere from 50 percent to 80 percent of the subjects (the actual proportion varied from study to study) yielded to the erroneous majority at least once, and overall, conformity occurred on over a third of all critical trials.

Milgram (1961) looked at the responses of adults who believed that they were being hired to test a new signaling system for jet airliners. In that context, the target subject was asked to judge the pitch of comparison tones relative to that of a standard tone.

“their judgment, but not mine, was distorted by the prospect of a large payoff instead of a small one,” or even that “they apparently thought it was worth playing a long shot; I don’t.” In other words, the introduction of the asymmetric matrix eliminated the most distinctive and potent feature of the Asch situation, which is the total absence of any suitable way for the naive subjects to explain the apparent discrepancy in perception.

Asch’s research represents one of the most stunning demonstrations we have of the remarkable power of situations to elicit behavior that most of us are sure we ourselves would never resort to – public conformity to the views expressed by others even when we privately hold utterly different views.

Darley and Latané undertook a number of studies to confirm the strongest version of this hypothesis – that is, that a victim’s chances of receiving help would be greater if there were only a single bystander available than if there were a whole group of such bystanders. In one study (Latané & Darley, 1968), male undergraduates at Columbia were left to fill out a questionnaire either by themselves, with two other subjects, or with two confederates of the experimenter instructed to remain impassive and continue working when the subsequent “emergency” occurred. This emergency consisted of a stream of “smoke” that began to pour into the room through a wall vent, eventually filling up the entire room. While 75 percent of the solitary bystanders left the room to report the smoke, only 10 percent of the bystanders who participated alongside the two impassive confederates, and only 38 percent of the three-person groups, ever intervened in this way.

Finally, in a New York University study (Darley & Latané, 1968), subjects heard someone whom they believed to be a fellow participant in an experiment feign an epileptic seizure while talking to them over an intercom system. When subjects believed they were the only listener, 85 percent intervened; when they believed that there was one other listener, 62 percent intervened, and when they believed that there were four other listeners, 31 percent intervened.

The consequences of this sort of dissonance reduction are revealed in Irving Janis’s (1982) well-known analysis of disastrous military and political decisions resulting from “groupthink.”

when asked to purchase another bond by someone who could sign them up on the spot, almost 60 percent put their names on the dotted line.

The principle in question, when applied by a skilled interpersonal manipulator, consists of first getting one’s “foot in the door,” that is, asking for a small favor or commitment (one that, in the context at hand, can scarcely be refused), and only then asking for the larger commitment or undertaking that constitutes the real objective.

Sliding scale.

People are prone not only to be influenced by situational factors but also to underestimate the extent of such influence (Nisbett & Ross, 1980; Ross, 1977).

Rather, what Milgram offered was a pointed reminder about the capacity of particular, relatively subtle situational forces to overcome people’s kinder dispositions.

Time and energy are saved, rumination and doubt are reduced, and nothing important is lost. But there is an obvious cost to our reliance upon the scripts, schemas, and other knowledge structures that help us interpret our world. When the cognitive representations we happen to choose or are led to employ turn out to be inaccurate in important respects, or when we employ them inappropriately (two almost inevitable problems whenever we venture into new social or intellectual terrain), the results are far less salutary. We are bound to make errors in interpretation or judgment, and we are apt to be slow both in recognizing that our preconceptions were wrong and in learning the lessons offered by our new experiences. Thus rapid and easy understanding as well as persistent, painful misunderstanding; warranted confidence as well as overconfident stubbornness; the capacity to be enlightened and informed as well as the capacity to be manipulated and misled are intimately related, indeed complementary, consequences of the tools that we all rely upon to construe our social environments. (See reviews by Cantor & Kihlstrom, 1987; Fiske & Taylor, 1990; Hamilton, Dugan, & Trolier, 1985; Markus & Zajonc, 1985; Nisbett & Ross, 1980; Petty & Cacioppo, 1985; Rumelhart, 1980.)

One particularly important set of biases constitutes the fundamental attribution error. This is the tendency for people to overlook situational causes of actions and outcomes in favor of dispositional ones. We will discuss these biases (Nisbett & Ross, 1980; Ross, 1977) later in this chapter and in great detail in Chapter 5.

people who thought they would sign the release also assumed that most ordinary people would do likewise and that the minority who failed to do so would probably be unusually shy or distrustful. People who thought they would not sign assumed that refusal would be the majority response and that the minority who agreed would be unusually gullible or exhibitionistic. This phenomenon is similar to one that has been called “egocentric attribution.”

There's a need for a personal application and awareness of this.

First, to predict a person’s response to a given situation – even a person whom one knows very well and has been able to observe in a wide variety of previous situations – one usually must know or correctly infer the details of that situation, in particular, those features of the context that help determine the relative attractiveness of the available response alternatives. Second, beyond knowing such objective features of the situation, one must discern the meaning of the situation from the private perspective of the actor.

Simply stated, no amount of reliability in the assessments of a single rater (nor even agreement between different raters) proves that the consistency lies in the behavior of the person being rated. A rater can persist in beliefs or stereotypes that are unsubstantiated by objective response data or that are substantiated only by interpreting such data in the light of one’s presuppositions.

Mischel and Peterson launched their challenging assault on conventional assumptions about personal consistency. Essentially, what these theorists did was to clear away the accumulated underbrush of studies that had relied exclusively on agreement of subjective assessments and to refocus attention on the few studies that had employed objective behavioral measures. In giving priority to these objective measures, Mischel and Peterson noted the obvious objections to the use of reliabilities in subjective assessment as a means of demonstrating the existence (much less establishing the magnitude) of enduring individual differences in personality. They insisted that while agreement of measures and stability over time are interesting phenomena that are worthy of investigation, they do not establish the validity of the trait constructs in question. They also summarized the existing literature on information-processing biases and the other shortcomings of human inference that could compromise the validity of social perceptions and even self-perceptions – in particular, reliance upon stereotypes based on appearance, role, or reputation. Moreover, the two decades that followed saw an avalanche of research on biases in social perception and cognition, providing an even stronger foundation for the critique (Dawes, 1988; Fiske & Taylor, 1990; Kahneman, Slovic, & Tversky, 1982; Nisbett & Ross, 1980; Taylor & Fiske, 1978). Many of the biases that were demonstrated were of just the sort tha...

Mischel (1974) and his co-investigators showed that any of several simple cognitive strategies that allowed children to divert their attention from the prospect of immediate reward could substantially enhance the capacity of virtually all children to delay gratification. In other words, such manipulations of context (and perhaps also of the meaning of that context to the children) could swamp the influence of any broad enduring differences in impulsivity, or patience, or any of the other individual differences among children that parents and professionals alike have in mind when they try to account for observed variability in the way different children respond to real-world opportunities and temptations.

It concerns understanding the situation nd less so about the trait

What does an individual’s distribution of responses across many situations look like in a world where people show the degree of behavioral consistency that Mischel and Epstein both seem willing to agree on? For example, how frequently would an “extreme” individual show extreme behavior and how frequently would such a person look rather average? Conversely, how frequently would an “average” person look extreme?

Good questions

For example, the scatterplot might portray the sociability of each member of the sixth grade in a particular elementary school measured in the lunchroom on a particular day as well as sociability measured for each child several days later on the playground. We see a great deal of variability in the amount of extroversion being manifested by the various children in each situation and a weak relationship between responses in the two situations – a relationship difficult to distinguish, at least by simple inspection, from no relationship at all (see Figure 4.1b). Obviously, a relationship as weak as this means that knowledge of any individual’s response in one situation gives us very little help in predicting that individual’s response in the other situation. To be more precise, the conventional regression formula tells us that the best predictions we could make in such circumstances would reduce our average or “standard” prediction error by only a very small amount (in fact, only about 1 percent). The reason for this is straightforward. When the outcomes being predicted are highly variable, as is the case for our sixth-graders’ displays of extroversion, the best predictions that we can make on the basis of knowledge of one previous response will, in virtually all cases, correspond to a level of extroversion very close to the average level manifested by the sixth-grade population in our study. In other words, the predictions we make generally will be similar to those we would...

To be sure, we would be able to predict very accurately the average level of shyness, aggression, or the like, that each child would manifest over a great many new observations. Indeed, we would be able to predict accurately each child’s entire distribution of responses. That is, we could predict confidently that each child’s distribution of future responses would closely resemble that child’s distribution of past responses. But we still would not be able to reduce by much our uncertainty about individual children’s behavior in individual situations. Specifically, knowing each child’s long-term mean or “true score” for aggression, and making individual predictions accordingly, would allow us to reduce the average discrepancy between what we predicted and what we observed by only about 8 percent. That is, our average error would be 8 percent smaller than if we simply guessed that each child’s behavior in each situation would correspond to the class average.

Contra peronalisism in a nutshell.

In short, a modest sampling of future behavior might very well allow us to see Harry (but almost certainly would not allow us to see Tom) do something truly extroverted.

Park (1986, 1989) found that while behaviors, affiliative memberships, attitudes, demographic information, and physical descriptions were all used with some frequency, traits (such as kind, shy, self-centered, easygoing) were more than twice as common as the next most frequent form of description. Ostrom (1975) asked college students to list the items of information they would want to know about another person in order to form an impression. Trait information accounted for 26 percent of all items listed; behavior, affiliative memberships, attitudes, and demographic and physical information together accounted for only 19 percent. Livesley and Bromley (1973) have shown that the use of trait terms increases steadily over the course of development for children in our culture, eventually becoming the most frequent type of description in free-response characterizations of other people. The meaning of the trait terms that people within a culture end up with would appear to be remarkably uniform.

Perhaps the most convincing line of work showing the layperson’s reliance on dispositional constructs of a trait type comes from a series of investigations by Winter and Uleman (1984; Winter, Uleman, & Cunniff, 1985), who showed that trait interpretations are made at the very moment behavior is observed and, in fact, may be integral to the coding of behavior.

And it further suggests that the dispositions they favor are suspiciously similar to the trait constructs fabled in song, story, and personology texts.

The most dramatic aspect of Figure 5.1 is the degree of consistency expected at the level of individual social behaviors. Subjects seemed to think the consistency from one situation to another would be far greater than research has ever shown it to be. They estimated the likelihood that the rank order would be preserved over two occasions was 78 percent – a likelihood requiring a correlation in the range of .80, whereas research suggests the actual correlation to be about .10.

In this spirit, Kunda and Nisbett asked professional psychologists, attending a convention symposium on statistical aspects of human judgment, to make the very same estimates as the college students whose data are shown in Figure 5.1. Some of these professionals would be describable as personologists, but most were social psychologists and experimentalists. The professional sample provided data that were essentially the same as those in Figure 5.1 – with one exception. The professional psychologists as a group were aware that traits are not very good predictors of behavior. (Or perhaps we should say they were reminded of that fact. Walter Mischel was seated prominently in the front of the room as they filled out their questionnaires!)

It is thus only for behaviors reflective of personality, and not for behaviors reflective of abilities, that people dramatically overestimate the amount of consistency to be expected and seem oblivious to the advantages of aggregated behavioral samples over individual instances. The latter point is particularly important because it suggests that people will make confident trait-based predictions on a small evidence base and will be unmotivated to increase their evidence base before making predictions.

A recent study by Brandon, Lawrence, Griffin, and Ross (1990) provided further evidence that laypeople expect levels of consistency and predictability in trait-relevant behavior that simply cannot be reconciled with the best available research evidence.

Their most important finding was their subjects’ willingness to predict that the individuals they had nominated would manifest very high levels of friendliness or shyness in any particular situation in which those individuals were examined. They further estimated that their nominees would behave in a markedly shy or a markedly friendly fashion far more often than they would behave in a relatively typical or average fashion. In other words, they made predictions that would be reasonable and appropriate if, but only if, behavioral consistency from one situation to the next yielded correlation coefficients not in the .10 range suggested by the objective empirical studies, not even correlations in the higher ranges sometimes claimed by critics of those studies, but, rather, correlations very close to 1.0.

The study indicates that observers are too willing to take behavior at face value, as reflecting a stable disposition (in this case, an attitudinal disposition), even when it is made abundantly clear that the actor’s behavior is under severe external constraints.

So not considering the context of the situation.

Attributing volunteering to a disposition rather than to the compensation offered. A study by Nisbett, Caputo, Legant, and Marecek (1973) showed that even such an obvious, widely appreciated situational factor as financial incentive can be slighted in explanation and prediction if there is a possibility of explaining behavior in dispositional terms.

Both contestants and observers rated the questioner as far more knowledgeable than either the contestant or the “average” student in the university.

Those who question are seen to be more knowledgeable.

People find it hard to penetrate beyond appearances to the role determinants of behavior, even when the random basis of role assignment and the particular prerogatives of particular roles are made abundantly clear. (Presumably, in everyday life, where such matters are more ambiguous, even less allowance would be made, and the behavior in question would even more willingly be taken at face value.)

there was no indication that subjects made much of a distinction between stability (which research often has found to be quite high) and consistency (which research has found to be almost uniformly low).

Sixty-three percent of subjects who were not in a hurry stopped to offer help to the “victim.” Only ten percent of subjects who were in a hurry offered help. In contrast, the dispositional measure concerning the nature of religious orientation played virtually no role in determining whether the subject stopped to help. The Darley and Batson experiment thus, in a sense, replicates but amends the lesson of the parable of the Good Samaritan. Their experiment invites us to surmise that all the priests and Levites who passed by on the other side of the road were simply running behind schedule!

The interpretation is not right.

Informing the subjects by having them read about the Darley and Batson study had no significant effect on their predictions about the effect of the dispositional variable of religious orientation. It did have an effect on their estimates of the effect of the situational variable of being in a hurry; but the effect was a mere 18-percentage-point difference, far less than the 53-point difference reported by Darley and Batson.

Astonishing.

A similar point was made in a study by Safer (1980) of students’ tendency to attribute obedience in the Milgram experiment to dispositional tendencies rather than to the power of the situation. Safer showed the Milgram obedience film to students. Despite the film’s emphasis on the extent to which it was the situational factors that compelled obedience, subjects substantially overestimated the amount of shock that would be administered when those factors were absent. Thus subjects continued to interpret behavior in terms of presumed dispositions, rather than recognizing the crucial role that Milgram’s particular situation had played in producing his disturbing demonstration.

Good.

nominators believed that their knowledge of peers’ reputations and personality permitted them to make confident predictions, at least about some individuals, and that their designated altruists and nonaltruists would manifest their dispositions regardless of the situational factors involved. But they were wrong! The situational variables proved more important than the relevant actors’ dispositions – more important, at least, than any dispositions salient to their peers. The evidence thus highlights serious flaws in some central tenets of lay personality theory. The evidence, in fact, is consistent with an extreme version of the fundamental attribution error. People readily make trait ascriptions from data that permit only a situational interpretation or, at most, the interpretation that the actor behaves in a particular way in a particular type of situation. These trait ascriptions are then used as the basis for yet further predictions, which, again, are characterized by little attention to situational factors.

For predictions of the kind studied by Dunning and colleagues, and by Newton and colleagues (where subjects were asked to predict who would be most likely and who least likely to contribute to a food drive), the base rate, whether known or presumed, is the best basis for prediction. When the base rate is extreme, one can oppose it in one’s predictions only at one’s dire peril. And this is true even if the target is someone who the predictor knows well.

Observers were less accurate than they believed, and they did not improve on their accuracy by adding their knowledge about the target to their presumed base rates for the situation, even though they believed that they had improved their accuracy by drawing on such knowledge. A little knowledge can be a dangerous thing, at least when it increases confidence far more than it increases accuracy. (See also Borgida & Nisbett, 1977; Nisbett & Borgida, 1975.)

Very telling. Just because you have knowledge does not mean you can use it well.