Chomsky's ultimate point is that we cannot build an adequate science of behavior (given the assumption that this is in fact what psychology is) simply on the description of the functional relations between stimuli and responses. Rather, we must develop a theory of the internal biological and cognitive structures, states, and organizations that mediate between stimuli and responses. There can be no doubt that understanding human behavior, or even most forms of behavior by non-humans when they are allowed to live in their natural environments as opposed to the behaviorist's laboratory, requires that we assume the existence of these factors and attempt to describe the principles by which they arise and operate. It should be obvious that such factors are attributes of the organism rather than properties of the situation outside the organism. Neither organismic nor environmental factors alone are sufficient to account for behavior: a proper explanation must take account of both in relationship to each other. As in psychology in general, so in the psychology of personality in particular. An adequate theory of personality must describe both kinds of attributes, and specify how they interact to produce behavior.
Appreciation of this state of affairs resulted, in the early 1970's, of the Doctrine of Interactionism (Bowers, 1973, p. 314). This states that "Both behavior and reinforcement are subject to selection by biocognitive structures". These structures include the biological substrates of mental processes; and the cognitive system which organizes them. Interactionists agree that a person's behavior is determined by the situation in which it occurs. But they also assert that the situation itself is largely determined by the person:
An interactionist or biocognitive view denies the primacy of either traits or situations in the determination of behavior; instead, it fully recognizes that whatever... effects do emerge will depend entirely upon the sample of settings and individuals under consideration.... More specifically, interactionism argues that situations are as much a function of the person as the person's behavior is a function of the situation (Bowers, 1973, p. 327; emphasis in the original).
There are many ways in which attributes of the person contribute to the determination of behavior. One is through biology, in the form of whatever individual differences in genetic and biochemical endowment exist prior to differences in the environments to which neonates are exposed. Setting such factors aside, there are two other modes of influence: cognitive transformations and behavioral selection. The person's cognitive apparatus selectively focuses attention, constructs percepts, reconstructs memories, and directs thought and action. The operation of these cognitive structures, or schemata, "filter and organise the environment in a fashion that makes it impossible ever to completely separate the environment from the person in it (Bowers, 1973, p. 328). Moreover, the attributes of the person, whether physical or behavioral, may literally alter the situation which he or she enters. Thus the person affects the situation by changing the pattern of dues, demands, expectations, and reinforcements -- and thus, of course, the behavior displayed by the individual exposed to them.
Early Interactionist Approaches to Personality
Interactionism was not an entirely new position within personality in 1973. In fact, several earlier theories were explicitly interactionist in nature (Ekehammar, 1974; Endler, 1982).
Murray: Interactions as Themata
Perhaps the first explicitly interactionist conception of personality was offered by Henry A. Murray (1938, 1951a, 1951b). In many ways, Murray's personology was highly influenced by Freud. For example, he adopted Freud's tripartite division of mental functions into the id, ego, and superego, and he agreed that behavior was primarily motivated by unconscious drives. However, Murray also differed from Freud in a variety of ways. For example, Murray held that the individual was driven by psychogenic needs such as achievement, affiliation, and dominance, as well as viscerogenic needs such as hunger, thirst, and sex. Table 11.1 lists Murray's (1938) psychogenic needs, along with brief definitions of them. Although the psychogenic needs were derived from the viscerogenic ones, they were no less important as determinants of behavior. Especially important, especially in the mature adult, were those having to do with understanding, mastery, and creativity. In stressing the role of psychosocial needs, then, Murray's theory had some of the character of the neofreudian approaches of A. Freud, Horney, Adler, and Fromm. In stressing the role of positive, "cognitive" needs, Murray anticipated somewhat the work of the psychoanalytic ego psychologists -- in fact one of the most prominent of these, Robert W. White, was Murray's protege.
Place Table 11.1 About Here
For Murray, as for all the dynamic psychologists, the process of understanding a person begins with understanding his or her needs -- the forces that motivate his or her behavior. However, Murray went beyond the other psychodynamic theorists in explicitly arguing that personologists must also consider those factors in the environment that support or prevent the expression of the individual's needs. Murray named these forces press (in Murray's terminology, the plural of "press" is "press"). Further, he distinguished distinguished between two kinds of press: alpha, relating to the environment as it is objectively described; and beta, relating to the environment as subjectively perceived and interpreted by the individual. Murray believed that beta press was much more important than alpha press, although he was especially interested in cases where there was a wide discrepancy between them -- a situation that Murray characterized as delusional. Table 11.2 shows Murray's (1938) list of common environmental press.
Place Table 11.2 About Here
In summary, Murray insisted that understanding persons entailed understanding both their psychological needs and the environmental press which impinged on them. He even coined a term, the thema, (plural, themata) to represent the particular combination of needs and press (particularly beta, or subjective, press) that characterized any episode in an individual's life. For Murray, themata find their expression in particular social interactions -- their constituent needs, press, and outcomes. Over successive episodes, simple themata combine into complex themata extending over the individual's entire life history. In addition, Murray employed the term unity thema to refer to that one combination of needs and press that best characterizes the individual's behavior. Formed in early childhood, Murray believed that this unity thema repeated itself, in one form or another, throughout the individual's life cycle. For Murray, the unity thema was the key to understanding individual experiences, thoughts, and actions that otherwise would be inexplicable.
Lewin's Formula and Its Implications
Interactionist concepts were also explicitly articulated in those personality theories which arose out of the Gestalt tradition in psychology, especially that of Lewin (1935, 1951). For Lewin, behavior was a function of the situation in which it was elicited. But he defined the situation as including both the psychological environment and the person. Following the Gestalt precept that the whole is greater than the sum of its parts, then, the situation is construed as a holistic entity in which the person and the environment depend on, and influence, each other. Expressed mathematically,
B = Behavior;
P = Person; and
E = Environment.
Let us consider the implications of such a formula, and see how interactionism works. For this purpose, consider dominant behavior, in which one individual controls or prevails over another individual or group, as when a person takes charge of some group endeavor or decision. According to Lewin's formula, there are two broad classes of determinants of dominant behavior: the general tendency of the person to behave in a dominant manner, or the general tendency of the situation to elicit dominant behaviors.
Consider first an extremely strong form of trait theory, which holds that behavior is wholly determined by pre-existing individual differences in in some personality traits. Such a claim may be expressed mathematically as follows:
The formula simply states that the probability or intensity of some behavior increases as a monotonic function of some attribute(s) of the person. Suppose that B stands for some behavioral index of dominance such as taking charge of a group, and P stands for subjects' scores on some psychometric instrument for measuring the trait of dominance. Thus, the formula expresses the prediction (or hypothesis) that the occurrence of dominant behavior in some situation will be determined by the person's trait of dominance.
This hypothesis might be tested in the following experiment. First, the investigator selects an instrument for the measurement of dominance. The trait might be assessed by self-reports, as in the Dominance subscale of the California Psychological Inventory, or by judges' ratings, as in the items pertaining to dominance in the California Q-Set. On the basis of such an assessment of personality, a representative sample of the population is divided into three groups: low, medium, and high in the trait of dominance. For the purposes of this example, we may arbitrarily assign the submissive group a value of 1 on the scale, and the dominant group a value of 1; the medium-scoring group, then, would be assigned a value of 2. Individuals from each group might then be placed in a group which was assigned to solve some problem. The investigator observes each subject's behavior, and records assertions of leadership and other manifestations of dominance in this standard situation. The value of B is predicted to be equal to the value of P.
The predicted results of such a study, assuming that trait theory were correct, would look something like those depicted in Figure 11.3. The horizontal axis represents the groups of low, medium, and high dominant subjects. The vertical axis represents the average number of dominant behaviors observed in each group. As would be expected, the submissive individuals show little tendency to take charge (showing only about 0 unit of dominance on average), the dominant individuals show a strong tendency in this direction (averaging 1 units of dominance), and the moderate individuals fall in between these two extremes (2 units). In the language of the analysis of variance, such an outcome is known as a main effect of persons.
Place Figure 11.3 About Here
Now consider an equally strong version of situationist theory, which holds that behavior is completely determined by the environment in which individuals find themselves. This notion may be given an analogous mathematical expression, as follows:
Again, the formula simply states that some behavior increases as a monotonic function of some attribute of the environment. Suppose that E stands for rankings of various environmental settings in terms of the likelihood with which people display dominance in them. Again, B stands for assertions of leadership and other dominant behaviors. Much as in the previous example, the formula expresses the prediction (or hypothesis) that the occurrence of dominant behavior in some environmental setting will be determined by the setting's tendency to elicit such behavior.
This hypothesis might be tested in the following experiment. First, the investigator must provide an assessment of environments analogous to the assessment of persons in the previous example. As noted in Chapter 10, however, general-purpose instruments of this type, analogous to the CPA or Q-Set, are not widely available. However, such an assessment might be made in the following manner. The investigator might ask a representative sample of subjects to list the settings in which they commonly find themselves. He or she then selects some number, say 10, that were commonly listed by the subjects, and thus likely to be encountered by almost everyone. A new sample of subjects then rates these settings in terms of the likelihood in which they would display dominant behavior in them. On the basis of ratings such as these, some settings are classified as low in environmental press for dominance, others medium, and still others high. As before, we assign values of 0, 1, and 2, respectively, to the situations in these categories. Finally, a new representative sample of subjects is unobtrusively observed in each situation, and the investigator records assertions of leadership and other dominant behaviors in each. The value of B is predicted to be equal to the value of E.
The predicted results of such a study, assuming that situationism were correct, would look something like those depicted in Figure 11.4. The horizontal axis represents the situations of low, medium, and high press press for dominance. The vertical axis represents the average number of dominant behaviors observed in category. As would be expected, subjects in the low-press settings show little tendency to take charge (showing only about 0 unit of dominance on average), those in high-press settings show a strong tendency in this direction (averaging 1 units of dominance), and those in moderate-press settings fall in between these two extremes (2 units). In the language of the analysis of variance, such an outcome is known as a main effect of environments.
Lace Figure 11.4 About Here
As presented here, the strong version of trait theory ignores the contribution of situations by averaging across stimuli, just as the strong version of situationist theory ignores the contribution of persons by averaging across subjects. These are of course straw men. In truth, each theory acknowledges the contribution of the other type of factor (though, to be accurate, it must be said that trait theorists are much more willing to concede the power of situations than situationists the power of traits). The question is,how do the factors combine in the joint determination of behavior? One possibility is that they act independently of each other. Such a relationship can be expressed in an additive model, as follows:
B =f(P + E).
The formula simply says that the effects of some attribute of the person, such as a trait of dominance, are added to those of some property the of the situation, such as its press for dominance, in the determination of some dominant behavior, such as exerting leadership.
Figure 11.5 shows dominant behavior plotted as a function of both the situation and the person, as predicted by the additive formula above. As in the trait example, three groups of subjects have been classified as low, medium, and high in general level of dominance, as measured by questionnaire. Moreover, as in the situationist example, three groups of settings have been classified as low, medium, and high in their press for dominant behavior. The subjects are observed in each type of situation, and their dominant behaviors recorded. In the figure, the vertical axis indicates the frequency with which each group of subjects displays dominant behaviors in each type of setting.
In Panel A, the subject groups have been arrayed along the horizontal axis, and the different types of settings are represented by separate lines. The predicted values of B may be found by summing the corresponding values of P and E. For the submissive group, where P = 0: 0 units of dominance are predicted in the low press situation (0 + 0); 1 unit in the medium-press situation (0 + 1); and 2 units in the high-press situation (0 + 2). For the moderate subjects, the corresponding values are 1, 2, and 3. For the dominant subjects, the corresponding values are 2, 3, and 4. Note that in both cases the relative differences between low-, medium-, and high-press settings is constant across the three groups of subjects.
In Panel B, the same results are portrayed differently. Here, the settings are arrayed along the horizontal axis, and the subject groups are represented by separate lines. Again, note that the relative differences between the low-, medium-, and high-dominant subjects is maintained in each class of settings. In the case of these examples, we have two main effects, one of persons and the other of environments. But there is not yet an interaction between these factors.
Place Figures 11.5a and 11.5b About Here
So what do interactions look like? Interactions can be expressed mathematically by a multiplicative rather than an additive relationship between two (or more) variables, as follows:
B =f(P x E).
In other words, P and E both affect B, but the effect of each variable depends on the level of the other. Predicted values of B may be obtained by multiplying the corresponding values of P and E. For example, where P and E both equal 0, then B also equals 0. Where P and E both equal 2, then B = 2x2 = 4.
The full range of values is plotted in Figure 11.6, for an experiment in which three subject groups differing in dominance have been observed in three classes of environmental settings that differ from each other in terms of their press for dominant behavior. Panel A arrays the subject groups along the horizontal axis. Comparing low-press to high-press situations, there is little difference in behavior among the submissive individuals; but for the most fearful ones, the situation to which they are observed makes a very big difference. Panel B resents the same results, except that the settings are arrayed along the horizontal axis. Comparing submissive and dominant individuals, there is little difference in behavior in the low-press situation; but in the high-press situation, the group differences are substantial indeed. In this case, where the behavioral effect of a trait depends on the environment in which that behavior is observed, and the behavioral effect of en environment situation depends on the kind of person who is in that environment, we have a case of the person by situation interaction.
Place Figures 11.6a and 11.6b About Here
In addition, personal and situational factors can also exert independent effects on behavior, a situation that might be expressed (without the weights) as follows:
BD =f(P + E + (P x E)).
Figure 11.7 shows the hypothetical values of B when P and E are both independent and interactive.
Place Figures 11.7a and 11.7b About Here
In the foregoing examples, we have assumed that personal and situational factors were equally powerful as determinants of dominant behavior. Of course, this is not necessarily the case. It might well be that traits are relatively powerful and situations relatively weak, or vice-versa. The kind of experiment that would be required to determine whether this is the case will be described later in this chapter. For the present, however, it is enough simply to state that differential determinative power can be represented in Lewin's formula by the addition of a weight, as follows:
B = f(bPP x bEE),
bP = the weight to be given to personality traits; and
bE = the weight to be given to environmental factors.
In this formula, the value of each factor, P and E, is multiplied by its respective weight.
Assume that the two weights of must sum to 1.0. Thus, in the case where P and E are equally potent, they both would receive weights of 0.5. Where P and E both equal 0, the corresponding value for B = (0.5(0)) x (0.5(0)) = 0; where P and E both equal 2, the corresponding value for B = (0.5(2)) x (0.5(2)) = 1. In the case where P is more powerful than E, the weights assigned might be 0.8 and 0.2, respectively; in the case where E is more powerful than P, the weights might be reversed. As indicated later in this chapter, there are reasons for thinking that the weight to be assigned to P factors is slightly larger than the weight to be assigned to E factors -- perhaps, on average, .55 vs. .45. Still, any interpretation of the differential size of the two main effects, for persons and for situations, must be qualified by the presence -- and strength -- of the person-by- situation interaction. Where such interactions are present, any statement about main effects -- regardless of what their differential weight may prove to be -- is likely to be highly misleading.
While the differential weight to be given to personal and situational factors may be of interest, it is important to note that the weights themselves have no impact on the form taken by the relations between P and E on the one hand, and B on the other. As another exercise, the reader may wish to confirm that the differences between groups remain constant across situations, and the differences between situations are still constant across groups. In other words, the relative values of B are not changed: all that is changed is the differential contribution of P and E factors to the final outcome. As an exercise, the reader may wish to determine that there are also main effects for persons (averaging across the three situations, dominant behavior increases as a function of dominance) and situations (averaging across persons, dominance increases as a function of environmental press). However, the presence of the person-by-situation interaction qualifies any statements that might be made about the differential power of persons and situations taken in isolation.
Forms of Interaction
In the foregoing examples, dominant behavior increased from low-press to high-press situations, regardless of the kind of person exposed to them. Personal factors only moderated that effect: dominance increased more slowly in the low-dominant individuals than in the high-dominant individuals (Figure 11.7a). In the same way dominant behavior increased from low-dominant to high-dominant individuals, regardless of the situation to which they were exposed. Situational factors only moderated that effect: the differences between the two groups were minimized in the low- press situation, and exacerbated in the high-press situation (Figure 11.7b). These kinds of interactions might be called fan effects, because the plot of B as a function of P and E looks something like an open fan.
However, other kinds of interactions also can be observed. Consider, for example, an alternative conceptualization of dominance and submissiveness. In the foregoing examples, we have construed submissiveness simply as the lack of dominance -- which is why submissive individuals were assigned values of 0 in the corresponding equations. Suppose, however, that submissiveness reflects the opposite of dominance, or the active suppression of dominant behavior in situations that call for it. This situation might be represented mathematically by creating a bipolar scale, and assigning negative values to highly submissive individuals while retaining positive values for highly dominant individuals. Lewin's formula for dominant behavior would remain the same, but with a value of -2 assigned to the submissive group the outcome on dominant behavior is quite different. Figure 11.8 shows that rather than increasing slowly across the various situations, dominant behavior by submissive individuals actually decreases when they move from low-press to high-press setting.
Place Figure 11.8 About Here
It is also possible for a complete crossover interaction to occur. Figure 11.9 depicts a hypothetical situation in which two groups of subjects, one high in dominance and the other high in submission, are exposed to two types of situations, one high in press for dominance and the other high in press for submission. In the situation that strongly demands dominance, the outcome is pretty much as we would predict: dominant individuals are dominant, while submissive individuals suppress this behavior. In the situation that strongly demands submission, however, the outcome is somewhat surprising: the highly submissive individuals actually display more dominance than the highly dominant counterparts. Such a result might be predicted by psychodynamic theories such as Freud's which permit unconscious motives to be transformed into their opposites.
Place Figure 11.9 About Here
Combining the two subject groups, however, as if one measured the response of the population at large without regard for relevant individual differences, the function would actually show no relation between threat value and response. By the same token, plotting dominance simply as a function of the personality variable without regard for situation effects would show no relation between situational variables and behavior. Put another way, such plots would actually misrepresent the determinants of dominant behavior.
Careful consideration of the variety of forms
that interactions can take demonstrates a point made by many
situationist critics of trait theory, and by trait-oriented
critics of situationist theory. That is, either position can
be vindicated by fortuitous selection of subjects and
situations (Bowers, 1973; Mischel, 1973). If, in the
preceding example, we had exposed subjects to two situations
both carrying demands for dominance but differing in the
intensity of that demand, we would never have observed the
crossover interaction. Put another way, where variance is
high on a personality attribute and low on a situational one
(as when extreme scorers on some questionnaire are exposed
to only a small range of situations), then the trait will
appear powerful; but where variance on the personality
attribute is low compared to that of a situational one (as
when unselected subjects are exposed to very distinct
situations), traits will appear impotent. From an
interactionist point of view, the trait versus situation
debate is misleading. But if in fact one view is more
correct than the other, the winner will be determined only
by studies in which there has been a representative sampling
of persons and situations.
Empirical Demonstrations of the Person-by-Situation Interaction
In principle, the person-by-situation
interaction is an appealing position, offering as it does a
kind of compromise between approaches that focus exclusively
on traits or on situations. Good ideas do not always work
out in practice, however, and so the value of the idea needs
to be put to empirical test.
Among the first to translate interactionist concepts into research was Cronbach (1957, 1975; Cronbach & Gleser, 1957; Cronbach & Snow, 19??). In his address as president of the American Psychological Association, Cronbach (1957) noted that psychologists could be divided into two camps, depending on their preferred methodologies.Experimental psychology is concerned with the effects of changing environmental conditions -- the effect of some independent variable on some dependent variable. In contrast,correlational psychology is concerned with the effects of the natural variation that occurs between individuals and groups -- the relation between two variables, predictor and criterion, neither of which has been deliberately manipulated by the experimenter.
In strict mathematical terms, the two types of design are equivalent in terms of the inferences that can be drawn concerning the relations between the variables in question. It was once argued that correlational designs did not permit inferences about causality. If two variables are correlated with each other, it is not necessarily the case that one is causally prior to the other. They both may be caused by some third, unmeasured variable. However, recent advances in statistical theory have yielded a new set of correlational techniques, such as cross-lagged panel analysis and structural equation modeling, that permit confident inferences about causality from correlational data. Still, the experiment has been the technique of choice for those who wish to study causal relations.
Be that as it may, the two designs have been differentially favored by various camps within psychology. At the time that Cronbach was writing, the prototypical experimentalists were the Skinnerian behaviorists, who observed (for example) the behavioral effects of different schedules of reinforcement. The prototypical correlationists were the ability testers, who observed (for example) the relations between IQ and scholastic achievement. In the present context, of course, the distinctions map directly onto the difference between situational and trait approaches to personality. Cronbach noted that each discipline tended to ignore the concerns of the other, and argued that for this reason neither could hope to yield a comprehensive understanding of psychological functioning. If so, then the solution is for the two disciplines to join forces to investigate the joint effects of the two types of variables -- those deliberately manipulated by the experimenter, and those that were not under experimental control.
Cronbach's argument in 1957 was largely theoretical, based on problems in student or personnel selection in education, business, and industry. The problem was construed as one of maximizing the payoff of the personnel decision. Given individuals differing in aptitude, what cutoff point should be instituted so that only those candidates most likely to succeed will be selected? Or, alternatively, given a number of different training regimes available, which one will lead to the best outcome? Consider, for example, the payoffs resulting when three different treatments are applied to people differing widely on some dimension of aptitude. One possible result is shown in Figure 11.8. Imagine that the subjects are high-school students differing in general scholastic aptitude, the three treatments are different ways on teaching typing, and payoff is the final grade in the course. Under Treatment A, high-aptitude students do much better than their low-aptitude counterparts; this relationship is reversed for Treatment B, where the low-aptitude students do better. With Treatment C there is very little difference in outcome along the dimension of aptitude, and as can be seen the average outcome is greater than for either of the other options. The situationist will assign all pupils to Treatment C, which maximizes the average expected outcome, and minimizes its variance within the population. Such a solution would be economically preferable, as the school board could hire just one typing teacher and purchase just one type of instructional material. From the point of view of instructional quality, however, a better solution would be to assign pupils differentially to instructional modes, depending on their ability levels. Low-ability students would do best with Treatment B, high-ability students best with Treatment A, and students of moderate ability with Treatment C.
Place Figure 11.8 About Here
In a later address, Cronbach (1975) reported the results of several studies in which interactionist concepts had been applied. For example, Domino (1968, 1971) assessed college students on two measures of achievement motivation: the motivation to achieve via conformity (Ac), and the motivation to achieve via independence (Ai). Students high in Ac got higher grades in classes where their instructors demanded conformity, while students with high scores on Ai got higher grades when their instructors pressed for independence. Table 11.1 shows the outcomes when independent and conforming students were taught by an instructor who demanded either independence or conformity. Exam scores, final grades, and student satisfaction were all maximized when the students were well-matched with their teachers.
Place Table 11.1 About Here
In support of his position, Cronbach (1975)
reviewed the results of 17 studies published in the Journal
of Personality and Social Psychology which had
analyzed the main effects of person and situational
variables (i.e., aptitudes and treatments) and their
interactions. On average, the interactions obtained were
approximately as large as the second-largest main effect. As
Cronbach argues, "If these main effects were worth
attention, so were the interactions" (p. 120).
Psychotherapy Outcome and the Patient-Therapy Match
A rather similar proposal has often been made with respect to the evaluation of success in psychotherapy. In an extremely provocative review, Eysenck (1952; see also Eysenck, 1966) proposed that insight- oriented psychotherapy of the kind practiced by psychoanalysts and other therapists of a psychodynamic persuasion was not effective in relieving the troubles of clinic patients. Examining the studies of therapy outcome available at the time, Eysenck noted that some patients, about a 2/3, got better after treatment. However, it is common in outcome research to include a control group of patients who are assigned to a waiting list, and thus do not receive any formal treatment. Eysenck discovered that the rate of improvement was approximately the same in the two groups. Apparently, active treatment by competent therapists -- which we can construe as a situational variable -- had no specific effect on mental illness.
Rather than concluding that mental patients were incurable, Eysenck proposed that an alternative form of treatment,behavior therapy, might be more successful. As indicated in the last chapter, Eysenck's intuitions appeared to be correct. In a landmark study, Wolpe (1958) claimed an astounding 98% success rate with a sample of phobic patients treated by systematic desensitization. Although Wolpe's figures may be inflated (Lazarus, 1985), there is no doubt that behavior therapy is very effective for a wide spectrum of mental illnesses (Rachman & Wilson, 1980). However, it was quickly pointed out that the kinds of patients typically treated with insight therapy are quite different from the kinds typically treated with behavior therapy. What was needed was a study that randomly assigned a representative sample of clinic patients to insight therapy, behavior therapy, and a waiting list. When such a study was performed, the results were surprising (Sloane, Staples, Cristol, Yorkston, & Whipple, 1975). In terms of short-term gains,both forms of active treatment were superior to the waiting list control, and the outcomes were equivalent in the two conditions; in terms of long-term gains, all three groups were equivalent. Again, when the different forms of therapy are conceptualized as situations, apparently there is no effect of treatment on outcome.
Leaving aside those who simply rejected such findings out of hand, response to findings such as these came in three forms. Reanalyses of the data by other investigators called into question Eysenck's precise figures. For example, Bergin (1971; Bergin & Lambert, 1978) concluded that the spontaneous recovery rate was somewhat inflated, and the recovery rate in the active treatment groups somewhat deflated, by certain methodological choices that Eysenck had made. Another group of investigators argued that certain ingredients were common to all successful therapeutic encounters, regardless of the professed ideology of the therapist (Frank, 1971; Lazarus, 1971, London, 1964, 1972). Thus, when insight therapy (or, for that matter, any other form of treatment) works, it works for the same reasons that behavior therapy works. When patients on the waiting list improve, it is because they find these active ingredients outside the formal therapeutic context. Stated as such, of course, this is merely a masked argument for a situational effect. However, a number of therapy researchers argued that the various forms of therapy actually differ, and that they interact with patient characteristics in complex ways (Bergin & Strupp, 1970; Fiske, Hunt, Luborsky, Orne, Parloff, Reiser, & Tuma, 1970; Garfield & Bergin, 1971; Luborsky, Auerbach, Chandler, Cohen, & Bachrach, 1971; Luborsky & Spence, 1978; Strupp & Bergin, 1969). Thus, when patients are randomly assigned to treatments, or relevant individual differences are not considered in the research design, no differential effect of treatment can be observed.
The obvious solution, of course, is for
research designs to take account of both person and
treatment effects on therapy, and the possible interactions
between them (Luborsky et al., 1971). Such studies are
extremely expensive to carry out, of course, so the
empirical literature on such interactions remains extremely
scanty. Nevertheless, some clinicians have already adopted
the spirit of interactionism when developing new forms of
therapy. For example, Lazarus (1973, 1985) has proposed that
the competent therapist is trained in a vast repertoire of
effective techniques, from which he or she selects those
that seem best suited to the patient's presenting complaint
and background. This multimodal therapy, which is at
least implicitly based on the idea of the aptitude by
treatment interaction, appears to be highly successful
The S-R Inventory
A second type of interactionist research makes use of a special type of questionnaire that poses a number of specific situations to subjects and then asks them to indicate the strength or likelihood of various responses within that situation (Endler, 1975, 1976, 1981, 1982, 1983, 1984; Endler & Hunt, 1966; Endler & Magnusson, 1976). The possible responses are all characteristic manifestations of some personality trait, and the situations are all contexts in which that trait might be displayed. In an S-R inventory of anxiety,for example, a subject might rate the extent to which she displays each of several responses (e.g., My heart beats faster; I sweat profusely) in each of several different situations (e.g., I am alone in the words at night; I am waiting in the dentist's office). In studies employing this technique, person variables are not assessed directly. Rather, each subject is considered to represent a different level of some relevant person variable. Table 11.2 shows some of the situations employed in an S-R Inventory of Dominance, while Table 11.3 shows some of the response modes which were assessed (Dworkin & Kihlstrom, 1978).
Place Tables 11.2 and 11.3 About Here
Such a questionnaire is completed by a representative sample of subjects, and their responses are subjected to a statistical technique known as within-subjects analysis of variance. Summing each subject's responses over all the situations, and over all the response modes, gives some idea of how generally dominant or submissive he or she is. Similarly, summing across all individuals and all response modes indicates how likely each of the situations is to elicit dominant behavior. And summing across all individuals and all situations indicates how potent each behavior is as a manifestation of dominance.
The same analysis of variance technique permits the assessment of the strength of the various interactions. The person-by-situation interaction, which is of most importance for present purposes, indicates the extent to which different individuals show different patterns of dominance across the situations sampled. The person-by-response mode interaction indicates the extent to which individuals differ in the ways in which they display dominance. And the situation-by-response mode interaction indicates the extent to which different types of situations elicit different sorts of dominant behaviors. These are known as two-way interactions, because they involve two factors. There is also a three-way interaction, person by situation by response mode, which indicates the extent to which individuals differ in exact manner in which they display dominance in various situations. For technical reasons, the three-way interaction is not always analyzed in the S-R Inventory technique.
Table 11.4 shows the results of the S-R Inventory of Dominance for male and female college students (Dworkin & Kihlstrom, 1978). As can be seen, individual differences among persons accounted for only about 10% of the variance in dominance manifested on the questionnaire; interestingly, differences among situations accounted for even less of the variance, approximately 8%. The interaction of persons and situations, however, accounted for approximately twice as much variance as either of these factors taken alone, about 22% (approximately 39% of the variance was unexplained by any of these factors, alone or in combination, as may be attributable to chance effects.
Place Table 11.4 About Here
On the S-R inventory, the person-by-situation interaction reflects individual differences in the pattern of behavior displayed across several situations. Several hypothetical patterns, as they might emerge from the S-R Inventory of Dominance, are depicted in Figure 11.9. The figure depicts the degree of dominant behavior (averaged across the 11 different response modes measured by the inventory) displayed by three subjects in each of the 12 different situations assessed by the inventory. Subject A is highly dominant in Situations 1-4, moderately dominant in Situations 5- 8, and not at all dominant in Situations 9-12. Subject B is highly dominant in Situations 5-8, while Subject C is highly dominant in Situations 9-12. As can be seen, the overall amount of dominance displayed by these individuals is roughly the same; so is the overall amount of dominance elicited by each situation. In terms of the analysis of variance, there are no main effects of person or of situation. However, each subject displays dominance in a different set of situations. In statistical terms, most of the variance in dominant behavior is accounted for by the interaction between persons and situations.
Place Figure 11.9 About Here
The S-R Inventory technique was introduced by Endler, Hunt, and their associates in 1962 (Endler, Hunt, & Rosenstein, 1962; Endler & Hunt, 1966, 1968, 1969). Since that time, the S-R format has been used in the study of a number of personality domains, including hostility, choice behavior, leisure activity, and aggression, as well as dominance. The general finding in these studies is that the proportion of behavioral variance accounted for by the person-situation interaction is considerably larger than that contributed by either of these factors taken alone (Bowers, 1973; Endler, 1973, 1975; but see Sarason, Smith, & Diener, 1975). For example, Bowers (1973), reviewing 19 comparisons in 11 different studies, found that the proportion of variance attributable to persons averaged 12.7%, to situations 10.2%, and to their interaction, 20.8%.
Of course, findings such as these are not
necessarily compelling. As has been pointed out, a
restricted sample of persons, situations, or response modes
could yield quite different results. For example, if the
situations sampled were all pretty much of the same type
(e.g., all involving friends), then the proportion of
variance attributable to situations, and to interactions
involving situations, would be correspondingly small
(Bowers, 1973; Mischel, 1973; Sarason, Smith, & Diener,
1975). However, to the extent that these questionnaires
sample a broad variety of situations and response modes, and
are administered to a representative sample of the
population at large, we may take values such as these to be
generalizable to the world at large (Dworkin &
Kihlstrom, 1978). The conclusion, that interactions account
for more of behavioral variance than main effects, is of
course congruent with the findings reviewed by Cronbach
The Template Solution
A final example of interactionist research involves an attempt to describe persons and situations in the same assessment language. Heretofore, persons and situations were described with entirely different sets of terms -- persons in terms of their traits, needs, and aptitudes; situations in terms of their physical and psychosocial characteristics and atmosphere. Bem and Funder (1978), however, proposed that the study of interactions would be facilitated if persons and situations were described with a common set of terms. At first glance, this would seem an unlikely prospect. Although Murray (1938) had made a start in this direction, situations are not easily described in terms of their achievement motivation, dominance, and anxiety; nor are persons easily described in terms of their average humidity levels, staffing ratios, and atmosphere. Bem and Funder proposed that situations be described in terms of the typical characteristics of individuals who behave in particular ways in them. After all, what is psychologically interesting about situations are the experience, thoughts, and actions of the people in them. In this way, situations are described in the same language that we would use to describe people.
For the purposes of a series of demonstration experiments, Bem and Funder (1978) employed the items of the California Q-Set as their descriptive language. In one study, children were observed in a delay of gratification situation. They were asked to express a preference for one of two kinds of snack foods (raisins or pretzels), and then were asked to wait for the snack, with the experimenter in the room, until they received permission to take it. Children who waited for 15 minutes were given their preferred snack; those who interrupted the experimenter before the appointed time were given their non-preferred snack. The children's delay times were correlated with each of the 100 items in the children's version of the Q-Set. The Q-Set items highly correlated (positively or negatively) with delay time, therefore, formed a description of the kind of child who can delay for a long time in that particular situation -- in other words, a description of the situation in terms of the characteristics of the persons who display a certain behavior in that situation.
Table 11.5 shows the Q-Set correlates of delay in the situation employed by Bem and Funder (1978). According to them, these items become a template against which other subjects could be measured in order to predict their behavior in that situation. The closer the child matches the characteristics of the average high-delay individual, the longer he or she would be expected to delay gratification. Interestingly, the template derived from Bem and Funder's subjects appears to differ considerably from those derived from experiments by investigators who employed other variations on the basic delay paradigm. For example, Mischel (1974) gave children a choice between two snacks, but the experimenter left the room during the waiting period; Block (1977) asked children to wait for an attractively wrapped package. Although the differences between these situations appear minor if not trivial, the fact that delay behavior shows different correlates in each situation indicates that a child who delayed in one would not delay in another. Very different children behave optimally in the three situations (for some limitations on these conclusions, see Mischel & Peake, 1982, 1983; also Bem, 1983; Funder, 1983).
Place Table 11.5 About Here
Rather than arguing for situational specificity, Bem and Funder argued that the differences among the templates were evidence for very subtle interactions between personal and situational characteristics. Apparently the various situations, while superficially similar, were functionally very different, as indicated by the fact that different kinds of children showed high levels of delay in them. In a later paper, Hoffman and Bem (1981, cited in Bem, 1983) further extended the template-matching technique. They noted that the Bem-Funder template-matching technique characterized persons in context-free, dispositional terms -- that is, what they were generally like as people. However, interactionism holds that such abstract descriptions are less useful than those that are more specific to a given situation. Accordingly, they proposed a technique of contextual template- matching intended to produce even finer-grained predictions of individual behavior in specific situations.
On the basis of other investigators' analyses of situational features, they developed a set of 150 descriptive items that could be applied to a wide variety of situations. This S-Set is analogous to the Q-Set of items used to describe persons. According to the procedure proposed by Hoffman and Bem, judges first rate a given situation with respect to the degree to which it possesses the various S-Set attributes. The smaller set of relevant S items then forms the basis for rating a particular person's personality. The judge considers each relevant S-Set item in turn, and then rates each Q-Set item in terms of the degree to which the person manifests that characteristic in that specific setting. The template is then formed from the Q-Set ratings, averaged across the relevant S-Set items. The person would be predicted to behave consistently across other situations that possessed the same features as the target situation. Similarly, other people who shared the template discovered in this way would be predicted to behave like the target individual in the target situation, and in other situations similar to it. However, no predictions could be made for persons, or for situations, with different sets of features. Behavior is determined by the interaction of particular personality characteristics with particular situation characteristics.
In this way, the unit of analysis in personality becomes what it should be, from the interactionist point of view: the person in context.
Hoffman and Bem (1981) did not actually try out
contextual template- matching in a classic prediction
situation. However, they were able to show that a variant on
their technique could lead to successful predictions of
behavior in a situation where the context-free
template-matching strategy had failed (Funder 1982).
Nevertheless, the results of these studies, taken together
with the research on aptitude-by-treatment interactions and
S-R inventories, are broadly consistent with the doctrine of
interactionism: personal and situational factors do not act
independently to determine behavior. Rather, they interact
with each other, and it is this interaction that determines
behavior. In predicting behavior, the units of analysis
cannot be just personality factors, or just situational
factors. The proper units of analysis have to do with the
In the classic prediction paradigm, causality is considered to be unidirectional. That is, something about persons, or about situations, causes a particular behavior to occur. The versions of interactionism discussed so far accept the unidirectional perspective, except that causal status is ascribed to the person-situation interaction, rather than to the person or the situation alone. There is another version of interactionism, however, that goes even further than this by asserting that each of the elements in Lewin's equation -- person, environment, and behavior -- exerts causal influence on the others. Not only do the person and environment affect each other, thus jointly determining behavior, but the behavior itself feeds back on its determinants to shape both the person who emitted it and the situation that elicited it. This is the doctrine of reciprocal determinism (Bandura, 1977, 1978, 1983a, 1983b, 1985). Reciprocal determinism may be graphically represented as in Figure 11.10.
Place Figure 11.10 About Here
The concept of reciprocal determinism embodies more than the idea of bidirectional causation -- that was already implicit in the notion that personal and environmental factors exert a causal influence on each other. Reciprocal determinism calls for a concept of holism rather than reductionism. As a scientific method, reductionism seeks to explain the properties of wholes in terms of the properties of their parts. In physical science, for example, the properties of compounds are explained in terms of their constituent atoms, and the properties of atoms are explained in terms of the properties of their constituent subatomic particles, etc. In biological science, the properties of organisms are explained in terms of the properties of their constituent tissues, and the properties of tissues are explained in terms of the properties of their constituent cells.
In social science, the properties of groups are explained in terms of the properties of their individual members; the properties of individual members, as physical and biological creatures, are explained in terms of their brain cells, hormone levels, neurotransmitter activity, etc. Always, causality is in one direction: from the elementary to the complex, in a single causal chain. In the case of psychology, reductionist explanation argues that causation proceeds from personal factors to behavior, and from situational factors to behavior; it may also permit situational factors to affect personal factors, or vice-versa. But it has no means of conceptualizing truly reciprocal causation, because no element is given priority over the others.
By contrast, in the dialectical view the properties of wholes emerge out of complex interactions among their elementary parts. Elements combine in such a way as to produce complex entities whose properties cannot be deduced from the properties of their parts -- in part because the larger entity changes those very properties. The dialectical method takes its name from the Greek word dialektikos, meaning "conversation". Its application to philosophy was described by Hegel in terms of three phases of investigation:thesis, in which an argument is posed;antithesis, in which its opposite is counterposed; and synthesis, a new proposition, different from both thesis and antithesis, and which resolves the conflict between them. The synthesis could not occur without the thesis and the antithesis, but it is quite different from them; and its existence changes them both by giving them new meaning.
Although the relevance of dialectical concepts
is widely recognized in the biological and social sciences
(e.g., Levins & Lewontin, 1985; Lewontin, Kamin, &
Rose, 1983), they have been difficult to put in practice.
This is simply because although we possess a vast repertoire
of quantitative techniques for the analysis of
unidirectional causation (e.g., Cohen & Cohen, 19??;
Judd & Kenney, 19??; Winer, 19??), we do not yet have
available comparable means suitable for the analysis of
reciprocal causal patterns. Some suitable statistics have
been developed (e.g., Thomas & Malone, 19??; other
Thomas papers), but as yet they have not been widely
applied. For the moment, we must be content with analyzing
separately the three pairs of bidirectional causal relations
and occasionally stringing larger pieces of them together,
and thereby gaining a glimpse of what the entire causal
structure would look like. These three causal pairs, P and
B, E and B, and P and E, may be called the three
The Dialectic Between Person and Behavior
Traditionally, psychology has been concerned with the internal structures and processes underlying behavior. Historically, these have been construed in many different ways: instincts, motives, traits, attitudes, beliefs, and values, for example. In each case, the idea is that behavior is the product of some internal state or disposition that is an attribute of the person -- however much it once may have been shaped by biological and social forces external to the individual. The idea will be familiar from the first two parts of this book. The psychometric approach to personality has been characterized by the search for, and measurement of, precisely these sorts of dispositions (a parallel effort can be seen in the classic attitude theories within social psychology). Similarly, the psychodynamic approach has been characterized by the attempt to uncover the motivational sources underlying behavior. Perhaps the notion of personal causation is expressed most clearly in theories like Eysenck's, who argues that introversion-extraversion, neuroticism, psychoticism, and intelligence are genetically determined; and Freud's who argued that sexual and aggressive instincts were part of our phylogenetic inheritance. But it is also expressed in any theory that assumes stability and consistency in behavior despite changing environmental circumstances.
The behavioral effects of many such internal structures and processes, limited as they are, have been documented in the chapters that have gone before. For illustrative purposes, we may select the correlates of self- control in the delay of gratification time. As discussed previously, delay of gratification studies typically require the subject (often, a young child) to defer choose between an immediate, less-preferred reward and a delayed, highly preferred reward. Many investigators (e.g., Bem & Funder, 1978; Block, 1977; Mischel, 1974) have employed the Q-Sort and similar techniques to reveal the kind of person who is able to delay gratification in such a procedure. Although the correlates of vary somewhat from study to study (Bem & Funder, 1978; Funder, Block, & Block, 1983; Mischel & Peake, 1983), a number of dispositional items appear to reliably correlate with delay. For example, the child appears to be attentive, expressive, competent, able to plan ahead, and capable of coping with stress. These qualities characterize the long-delaying subject as a child, close to the time that delay behavior is experimentally assessed, and as an adolescent, long after the test has been forgotten. In large part, they are the qualities that we would expect to find in a mature person, of any age, who possesses a high level of ego development.
What has not been discussed, up to now, are the reciprocal effects that behavior has on the person who emits it. One example of how behavior affects the person is provided by the self-perception theory of attitude change. Bem (1967, 1972) proposed that the person's perception of his or her own behavior leads to the formation of an attitude that is consistent with that behavior. This statement reverses the usual direction of causality: it is not attitudes that cause behavior, but rather behavior that causes attitudes. The theory assumes that we have little or no direct introspective awareness of our own internal states and dispositions (see also Nisbett & Wilson, 1977), and that our statements about such matters really reflect inferences about ourselves -- inferences that other people would also make if they observed our behavior, even though they cannot read our minds.
Self-perception processes were demonstrated in an experiment by Salancik and Conway (1975) in the area of religious attitudes. As part of a survey, subjects were asked to describe their own religious behavior -- whether they read religious literature, attended services, and the like. Half of the subjects were asked whether they did these things "occasionally", while the remainder were asked if they did these things "frequently". Most subjects in the sample engaged in such activities occasionally, but hardly any did so frequently. Thus, the subjects in the "frequently" condition were made aware, by their negative responses, that they did not engage in much religious behavior. Subjects were randomly assigned to the two conditions, so we may assume that their actual religious attitudes and activities were equivalent at the outset of the experiment. Nevertheless, a post-test questionnaire indicated that students who reported many occasional activities expressed stronger religious attitudes than those who denied many frequent activities. Thus, awareness of behavior seems to affect one's attitudes.
A more revealing study was performed by Chaiken and Baldwin (1981) on environmental attitudes. On the basis of a questionnaire, subjects were classified with respect to their attitudes toward environmental conservation. Most subjects express pro-conservation attitudes, so the subjects were classified in terms of whether their attitudes were strong and consistent, or weak and inconsistent. The investigators then induced the subjects to reflect on their behavior: half the subjects in each group were asked to whether they engaged in certain relevant behaviors "on occasion"; the others were asked whether they engaged in the same behaviors "frequently". Most subjects, if they are being truthful, will agree that they litter occasionally: therefore their anti-environmental behavior was made salient to these subjects. Most subjects, if they are being truthful, will also deny that they litter frequently: they are made aware of their pro-environmental behavior. When the subjects' attitudes were remeasured, it was found that those who were induced to report on their anti- environmental behavior were less strongly pro-conservation than those who had been induced to report on their pro-environmental behavior. However, this was the case only for those who had weak and inconsistent attitudes to begin with. Those who had initially expressed strong, consistent pro- conservation attitudes did not respond to the manipulation , and their post-test scores fell between those of the other two groups. Thus, awareness of one's behavior can create an attitude where none existed before. However, behavior had little effect in cases where the person already had a strong attitude. Thus, person variables (whether an attitude is present or not) moderate the effect of behavior on the person.
A similar sort of process may be at work in the "foot-in-the-door" effect on compliance (Freedman & Fraser, 1966). Our intuitions tell us that people are likely to agree to large requests if they have first agreed to small requests. In part, the reason has to do with item difficulty: subjects are not likely to "pass" a difficult item if they "fail" an easy one. But there is more to the phenomenon than this. In an experiment, homeowners were contacted door-to-door and asked to sign a petition in support of a local safe-driving campaign. No monetary donations were requested, and so it is not surprising that virtually everyone agreed to sign. Sometime later, the homeowners were recontacted and asked to place a large, unattractive poster in their yards supporting the campaign. Although the compliance rate was relatively low, those who had previously signed the petition were significantly more likely to comply than other homeowners in the same area, who had not been asked to sign the petition: 55% vs. 17%.
Similar effects have been obtained by others
(Pliner, Heather, Kohl, & Saari, 1974; Snyder &
Cunningham, 1975). Interestingly, it has been shown that
compliance is increased when the first request is relatively
large (Seligman, Bush, & Kirsch, 1975) -- although it
probably cannot be too large, or else people will not
complying in the first place, and the petitioner will never
get his or her foot in the door. Presumably the homeowners
would have agreed, if only they had been asked, so the
difference in behavior is not merely a matter of item
difficulty. Rather, it appears that the initial behavior,
created or strengthened an attitude that had not been
present before. This new attitude made subsequent
attitude-relevant behavior more likely than it was
previously. Here we have a case of both sequential and
reciprocal determination: an environmental event (asking a
small favor) leads to behavior that creates an attitude (a
personal factor) that reciprocally causes a behavior
(compliance with a large request).
The Dialectic Between Environment and Behavior
Although the functionalist psychology of James, Dewey, Angell, and others was concerned with mind in context, the external structures and processes governing behavior were initially the province of sociology. From the beginning, sociologists have studied, among many other things, the effects of social-structural variables such as race, class, and sex on behavior, as well as the effects of a person's role and status within society on his or her behavior. Social psychology, always closely linked with sociology, shared its interest in these matters, as well as a concern with the effects of the presence and behavior of other people (Allport, 1968; Jones, 1985). Classical experimental social psychology was very much a psychology of the situation -- as indicated, for example, in the random assignment of subjects to treatment conditions and the general failure to include in most designs any assessment of internal variables. While sociology and social psychology were chiefly interested in the behavioral effects of the sociocultural milieu, environmental or ecological psychology have pointed out the corresponding effects of the physical environment. The behaviorist revolution of Watson and (later) Skinner, of course, pushed doctrinal situationism to the farthest degree. It can be argued that for many social psychologists, environmental variables were employed as "stand- ins" for internal variables -- as, for example, Festinger's (1957) manipulation of the experimental situation in order to create cognitive dissonance. The radical behaviorists, however, abjured any interest in internal processes, and focused exclusively on the part played by events in the world outside the behaving organism.
Many apparent effects of environmental variables have been discussed in the previous chapter. However, as the critique of situationism showed, these effects are actually somewhat ambiguous because in so many instances environmental stimuli are filtered through the perceptual-cognitive system -- structures and processes that are rightly regarded as properties of the person. Thus, speaking strictly, the causal route is from the environmental event (an external factor) to the mental representation of that event (an internal factor) to behavior. In other words, it is not clear if these effects of the situation on behavior are direct, unmediated by internal structures and processes. Of course, from the point of view of reciprocal determinism, there is no problem if many situational effects exert their causal effects through mediation by personal factors. However, the elegance of Figure 11.10 seems to demand that at least some effects of E on B be relatively direct (for the same reason, we will encounter this problem again when we consider the effects of on P). Such effects may be found in the taxes and instincts of lower organisms. There, behavior is an automatic, obligatory response to environmental conditions. Of course, the stimulus-response relations are mediated by the organism's nervous system, and the instinct itself is a reflection of its phylogenetic endowment, but these are trivial objections. Taxes and instincts are universal within a species: every individual species member engages in those behaviors under the specified conditions. Thus the organism's individual history, current state, and future goals are entirely irrelevant. For all intents and purposes, then, the organism is irrelevant, too. Human beings do not have identifiable taxes and instincts. While the point with respect to instincts might be argued by sociobiologists (Dawkins, 19??; Konner, 19??; Wilson, 19??), this is not the place to carry on that argument (see, however, Gould, 1983; Levins & Lewontin, 1985; Lewontin, Kamin, & Rose, 1983).
Humans have reflexes, however, and these may be thought of as relatively pure responses to the situation -- after all, they occur even in the absence of a functioning cortex -- but they are not the stuff of which complex social behavior is made. More to the point are such things as the generally disruptive effect of noise, the generally facilitative effects of the presence others, perhaps the influence of pain and frustration on aggression, and the induction of sexual arousal and other emotional states by certain stimuli. Here again, the effects seem quite general, so that for analytic purposes the details of the individual's personality can be considered to be more or less irrelevant to the outcome. Somewhat more difficult to classify are "mindless" responses to environmental events that occur automatically, and without any deliberate thought on the part of the person (e.g., Langer, Blank, & Chanowitz, 1978). Again, to the extent that such responses are fairly universal, it would seem that the attributes of the particular person are fairly irrelevant to his or her behavior, and therefore that for all intents and purposes the effect of environmental variables is direct.
Following the lines of a previous example, we may look to delay of gratification for an illustration of the "pure" effect of environmental variables on behavior. Consider, again, the child who is asked to wait for a highly desirable treat: if he or she cannot tolerate the delay until the experimenter returns of his or her own accord, the child will receive something that is considerably less rewarding. If neither promised reward is present in the room, the average child can wait much longer than if either one is in sight; if both are in sight, the average child cannot wait very long at all. Similarly, if the rewards are physically present in the room, but hidden from sight, the child can delay gratification much longer than if he or she has them continually in view. Thus, just as certain personal attributes affect the child's behavior in the delay paradigm, so it is affected by the features of the environment to which he or she is exposed.
As seen in the experiments on self-perception, situational effects on behavior can have important subsequent effects on the person. This is not to say that internal factors cannot modify the behavioral effects of such environmental stimuli -- they certainly can -- only that internal factors of the sort labelled as P are not necessarily involved in these reactions. In line with the doctrine of reciprocal determinism, however, it is also important to consider the effect of behavior on the situation. In some sense, examples of such effects are too obvious for the point to require comment. After all, Skinner gave an important form of learning the name of operant conditioning precisely because the responses in question changed the environment that elicited them. And obviously, the behaviors of aggressive or dominant individuals have clear and sometimes dramatic effects on the environments in which those people live.
For an example, let us return to delay of gratification in children. Even when the rewards are clearly in sight, however, the child can engage in behavior that can transform the situation from one in which delay is unlikely to one in which it is highly possible. In the reward-present conditions, Mischel and his colleagues have often observed the children engaging in behaviors that effectively distract themselves from the objects in question. They may avoid looking at the rewards, or cover their eyes, or put their heads down on a table. They may also talk or sing to themselves, or invent little games to keep them occupied. Through their behavior, of course, they are effectively altering the situation from one in which the reward is present to one in which it is not. In so doing, they are creating, through their behavior, very much the kind of situation in which long delays are possible. Observations such as these lead us to consider the reciprocal effect of behavior on situations.
Even more interesting, perhaps, are instances where behavioral effects on the environment reverberate to affect the person, or future behavior, or both. Consider, for example, research on situations choice by subjects differing in personality. There exists a large literature (reviewed by Snyder & Ickes, 1985) showing that individual differences in personality traits influence the situations that people choose to enter and the activities in which they choose to engage. For example, Furnham (1981) has reported that extraverted individuals are more likely than their introverted counterparts to select situations involving assertiveness, competitiveness, and intimacy. Diener and Emmons (19??) prepared a list of situations common to a university environment (e.g., those involving one's living unit or other social group, stimulus seeking, cultural, and team or individual sports. Table 11.6 shows examples of each of these categories. They also administered a personality questionnaire to a large sample of college students, and correlated the scores with the students' preferences for each type of situation. Subjects high in aggression and stimulus- seeking preferred situations involving individual or team sports; those high in autonomy preferred cultural situations, and individual but not team sports; highly intellectual subjects preferred cultural situations above all others.
Place Table 11.6 About Here
In these studies we observe cases of the behavioral selection of situations. In other words, individuals behave in such a manner as to place themselves in some situations but not in others. In effect, then, the person's behavior has changed the character of the situation. The choices of these individuals are interesting because each choice makes certain behaviors relatively likely, and other behaviors relatively unlikely -- yet another instance of bidirectional causation. Thus, aggressive, competitive individuals who engage in many team athletic contests will have the opportunity to observe themselves engaging in aggressive, competitive behaviors -- behaviors that may serve, through self-perception principles, to strengthen relevant attitudes and dispositions.
We also observe, as Allport (1937) suggested we
might, a congruity between the attributes of the individual
and the kind of situation that he or she selects or
constructs behaviorally. Thus, the choice behavior is, at
least in some immediate sense, determined by personal
dispositions; these behaviors alter the situation in which
the person exists; the situation tends to increase the
likelihood of the very behaviors that the person is already
disposed to display; and the occurrence of these behaviors
is likely to strengthen the disposition itself. This pattern
of reciprocal and sequential causation. This brings us full
circle, and leads us to return to the beginnings of the
doctrine of interactionism, and to examine the relations
between persons and environments.
The Dialectic Between Person and Environment
The phenomenon of interpersonal attraction -- whether one person likes another -- is as good a place as any to begin a discussion of environmental affects on the person. Liking is an attitude, a generalized tendency to evaluate another person positively (such is the way social scientists write about love!), and as such is an internal, personal determinant of behavior. Obviously liking is affected by characteristics of the target: we tend to like those who are physically attractive, competent (especially if their competence is moderated by an occasional small blunder), and similar to us (the notion that opposites attract is largely a myth). Because the target is part of the world outside the person, these might well be considered environmental effects on liking. But there are also other situational effects at play, such as proximity. All things being equal, the likelihood that two people will develop a friendship is a function of the distance between them. Studies of new suburban tracts, to which everyone moved at approximately the same time and in which everyone is similar in socioeconomic status, show that most people who live near each other consider each other friends, and that few people have friends who live far away. The same holds within a living unit such as a dormitory, or a work unit such as a classroom. Many readers of this book developed firm friendships with their freshman year roommate, and failed to do so with an even more likeable person who lived down the hall, or on another floor, or in another building. Similarly, many college friendships develop simply because two people happened to sit next to each other in class.
Certainly the effects of proximity on liking make sense: people who live and work near us are simply more available to us as friends. However, more is involved than mere availability, because the average person has more friends among his or her neighbors than would be expected by chance. In part, the effects of proximity are mediated by familiarity: repeated exposure to an object increases judgments of likeability -- even if there is no substantive contact between the perceiver and the object. The effect of familiarity has been demonstrated in a clever study of photograph ratings. Subjects were asked to evaluate photos of their own and their friends' faces. Half the photographs reversed the left and right sides. The judges preferred the original photos of their friends, which depicted them as they would appear to an observer; however, they preferred the reversed photos of themselves, which depicted them as they appeared to themselves in a mirror. This is a pure situational effect: people preferred the view that they more frequently encountered.
The effects of mere exposure can also be observed in instances of observational learning (Bandura, 1965, 1977, 1985; Bandura & Walters, 1963). As noted in the previous chapter, behaviorist approaches to learning argued that new stimulus-response associations were formed through the direct experience of response consequences. Given the multitude of possible responses to a stimulus, those that are reinforced are become more likely to be emitted on subsequent occasions than those that are not. However, classic associationism held that no learning could occur in the absence of reinforcement. As indicated, experiments such as those of Tolman (1949) showed quite clearly that learning could occur without any apparent reinforcement to the behaving organism. Rats could learn the pathway through a maze simply by exploring it, without being rewarded for any specific turns.
In the case of human beings -- and, to a limited extent, non-humans as well (Mineka, 1985) -- most learning appears to be of that type. We pick up large amounts of knowledge about our world through precept and example, without any direct experience of what we observe and what we are told. That is what happens, for example, when we learn something from a book or a lecture. Aspects of personality can be acquired in the same way. Thus, children can acquire aggressive behaviors, and learn to direct them toward certain objects, simply by observing the behavior of other children, or adults, or characters in the movies or on television. The same is true for more generalized, enduring features of personality. For example, a major determinant of self-efficacy -- people's sense of themselves as competent -- is what they have been told about themselves, and what they observe in others who are similar to them in other respects. To the extent that the source of self-efficacy lies in the world outside the person, it may be thought of as an effect of the situation upon the person.
Another kind of situational effect can also be observed in studies of intrinsic motivation -- the person's desire to engage in some activity, without any promise or prospect of reward. Many people paint portraits, run marathons, and collect stamps because it is profitable for them to do so: they gain fame or wealth from the activity. Others of us, however, engage in the same sorts of activities but get nothing for it. Although a psychodynamic psychologist might wish to speculate on the deep, pre-oedipal sources of such behavior, most people are satisfied with the explanation that, for example, "She does it simply because she enjoys it". Such behavior is intrinsically motivated, and as such would properly be discussed under the dialectic between person and behavior. However, a classic study by Lepper, Greene, and Nisbett (1973) indicates that situational factors can undermine intrinsic motivation.
In their study, nursery-school children were engaged in an enjoyable task, drawing on large sheets of paper with felt-tip pens. This is the kind of behavior that youngsters will engage in for long periods of time without any apparent loss of interest. By any standard it is intrinsically motivating. In the Lepper et al. study, half of the children were promised a "Good Player Reward" for drawing with the pens; others were not promised anything, and then both groups were given time to draw. Later, the children were observed in a free-play period, in which they had an opportunity to resume drawing, but also were able to select other activities. Somewhat surprisingly, the children who had been rewarded for drawing now spent less time in this activity than those who had not been rewarded. Apparently, the reward undermined the children's intrinsic motivation to draw. This we may take as a situational effect on the person.
Later studies by Harackiewicz and her colleagues have clarified the effects of the reward. These studies have employed various populations and tasks and rewards, from schoolchildren solving interesting puzzles in order to receive ice-cream cones to college students playing pinball in order to receive movie passes. Some subjects are not rewarded at all; others are rewarded simply for engaging in the activities; still others are rewarded only if they meet or exceed a certain standard of performance. Later observations in free-choice situations reveal that subjects who have been rewarded merely for task-completion lose interest; by contrast, those who have been rewarded for achievement maintain or even increase their motivation for the task. Still, a situational variable -- reward structure -- affects the an internal attribute of the person: his or her motivation to engage in some activity. In addition, certain attributes of the person -- such as general achievement motivation -- serve to moderate the effects of reward structure on his or her motivation for some specific activity. This leads us to examine the reciprocal effect of the person on the environment.
Finally, what about the reciprocal effects of the person on the environment? Obviously, people affect the environment through their overt behavior, either by selecting one situation over another or by changing the features of the situation in which they find themselves. But is there any way that the person can affect the environment independent of overt action? On reflection, there are two possible mechanisms for this effect.
First, it is important to recall that the environment includes other people. Research on the perception of situations, to be discussed in detail later, clearly shows that other people, their characteristics, and their actions are important features of the actor's environment -- they are part of the context in which his or her behavior takes place. We have already seen the importance of other people for the actor and his or her behavior in the effect of the presence of others on helping behavior, modelling effects on aggression, and exposure effects on liking. In the same way that other people are part of the actor's stimulus situation, the actor is part of the stimulus situation for other people, and may alter their behavior in important ways. Thus, the person's presence may change his or her own environment, even though he or she is merely present, and takes no other action.
This kind of effect may be illustrated by the development of gender roles -- a topic discussed later in more detail. Every culture has certain normative standards for the behavior of males and females. In Western culture, for example, the stereotype of masculinity is agency: boys and men are expected to be active, competitive, and independent. In contrast, the Western cultural stereotype of femininity is communality: girls and women are expected to be caring, sensitive, and emotionally expressive. These expectations are so deeply ingrained in our culture that we often take them to be natural, and assume that they are related to genetic, biochemical, and structural differences between the sexes. In fact, studies of gender dimorphism (e.g., Money & Ehrhardt, 1972) reveal that biological factors are clearly important to the process. Genetic messages from the sex chromosomes control the differentiation of the internal and external reproductive systems. In addition to these effects on the individual's physical characteristics, there also appear to be small but significant effects of the sex hormones on such behavioral characteristics as activity level.
But the same studies also reveal the importance of sociocultural factors. Once the sex of the child has been identified, the parents and other agents of socialization begin to impose different socialization regimes on male and female children -- learning experiences carefully designed to foster the development of those patterns of experience, thought, and action that the culture deems appropriate for males and females. Evidence for the importance of sociocultural factors comes from three sources. First, there are observational studies of the different ways that boys and girls are treated by parents and others during the first five or six years of life, when gender identity, gender role, and perhaps even sexual preference are established. Second, there are anthropological studies that reveal large differences in gender roles across cultures, despite the fact that the biological differences between the sexes are constant. Finally, there are clinical studies of sex-reassignment. For example, there are cases in which genetically and hormonally normal boys suffer trauma to their external genitalia, and therefore are reassigned to be brought up as girls. After surgical correction of their external physique, these children typically develop "normally", in accordance with the cultural standards imposed on them, despite the fact that they are carrying the genetic and biochemical endowment of males. Although all cultures make some distinctions between the genders, and each individual develops some sense of gender identity, there appears to be nothing that is obligatory about the content of gender roles except the "procreative imperatives" of menstruation, impregnation, gestation, and lactation. Despite the familiarity of masculine agency and feminine communality, it appears that these characteristics are largely imposed by the culture, rather than the individual's biological endowment.
Despite the clear role of experience in the development of gender- role, it should be clear that the entire socialization process is structured by the physical appearance of the child. In normal development, the parents take notice of the child's external genitalia -- whether there is a penis and scrotum or vagina and clitoris -- and then create socialization experiences deemed appropriate to the child's gender. In the cases of sex-reassignment, the appearance of the genitalia changes and socialization pressures change accordingly. In both cases, the effect on the environment of the child's genitalia are strong and pervasive. Almost everything is different: the colors of the walls, the clothes (even the colors of the clothes), the toys, the amount of close touching and roughhousing -- these are just a few of the common differences in the environments in which even infant males and females are exposed. This differences are imposed by the culture, to be sure, and thus classify as effects of the environment on the person. But the course of the cultural difference is in the physical appearance of the person. Thus we see the reciprocal effect of the person on the environment in which he or she lives -- even though the person him- or herself has done nothing at all.
Clues to another way the person affects the environment come from recognition of the fact that the environment has few direct effects on the person. Even in the case of ostensibly "mindless" behavior, and in observational learning, the effects of the environment are mediated by the individual's cognitive processes. Murray (1938), one of the first personality psychologists to take serious account of the role of the environment, clearly underscored this point by distinguishing between alpha press (the objective environment) and beta press (the subjective environment), and in assigning priority to the latter. In larger part, what determines behavior is the person's mental representation of the environment, rather than the environment per se. The environment that elicits behavior is literally constructed through perception, categorization, and inference, and reconstructed on subsequent occasions through memory. The individual's ability, through active cognitive processing, to transform the his or her mental representation of the world -- to literally perceive the world differently -- is another way in which the person can affect the environment, independent of overt behavior.
The effects of cognitive transformations are dramatically illustrated, again, in studies of self-control in young children. As noted earlier, some children have the capacity to delay gratification for long periods of time (an effect of the person on behavior), and some situations permit longer delays than others (an effect of the environment on behavior). As a general rule, situations in which the promised rewards are absent, or at least hidden from sight, appear to be associated with delay times. But as we also observed, children can act on their environments in such a way as to transform them from ones in which delay is extremely difficult to ones in which it is relatively easy. Even when such a behavioral transformation of the environment is not possible, even fairly young children can still act to transform their mental representations of the situation that faces them (Mischel, 1974, 1981, 1984). Thus, children can increase their delay times simply by not thinking about the rewards. Contrary to psychoanalytic notions of hallucinatory wish-fulfillment, thinking about desired objects does not increase the child's ability to wait for them. Thinking about pleasant topics also increases delay times, while thinking about unpleasant topics impairs delay of gratification. Children who focus on consummatory aspects of the rewards, such as their taste and texture, cannot delay for very long. However, children who transform pretzels into toy logs and marshmallows into clouds can wait for extremely long times. Children who imagine that the real rewards are only pictures delay longer than those who imagine that pictured rewards are the real things. As they develop, children become increasingly aware of the utility of these strategies, and increasingly good at deploying them effectively (Mischel & Mischel, 1983).
Similar kinds of cognitive strategies have been found to be effective in reducing stress and pain. Thus, surgical patients who catastrophize -- who focus on the unpleasant aspects of their situation -- show far more signs of stress, and experience much more pain, than those who redefine the situation in positive ways, focusing on pleasant aspects even if they are far in the future (Janis, 19??; Leventhal, 19??; Michenbaum & Turk, 19??).
Studies of these begin to answer the central
question in interactionism: what is it about the person that
affects the situation? Certainly personality traits like
ego-control affect behavioral outcomes such as tolerance of
frustration, but how do they affect the environment? In
Mischel's studies of self-control we begin to see what lies
behind the personality traits of long-delaying children:
they possess a set of cognitive and behavioral skills, and
associated knowledge of themselves and the world, that are
lacking in children who cannot tolerate long periods of
frustration. The attributes of the person that are most
important to the person-situation interaction, then, are
cognitive in nature: the person's fund of knowledge about
him- or herself and the social world outside the self, and
the person's repertoire of skills and strategies by which
that knowledge can be used to plan and execute competent
behavior intended to achieve the person's goals and cope
with environmental events. In the remainder of this book, we
focus on these social-cognitive aspects of personality, as
an alternative to the dispositional and motivational aspects
emphasized by the psychometric and psychodynamic theories.
The Social Interaction Cycle
The dialectical idea of reciprocal causal processes, and the role of cognitive factors in the person-situation interaction, is most completely illustrated in the general social interaction cycle, as depicted in Figure 11.11. Every social interaction consists of at least two participants. Each person is simultaneously an actor and the target of the other's actions. However, for purposes of simplicity we will arbitrarily identify one participant as the actor and the other as a target. (Of course, in social interactions where the two participants are unequal in status, these designations are less arbitrary. In this illustration we assume that they are roughly equal.) It is commonly held that all interactions are characterized by social exchange: each participant wants something of the other, and the participants negotiate in order to achieve the best possible outcome for themselves (of course, in cases of highly unequal status, the best possible outcome for one will be simply to minimize losses). Therefore, each participant enters the interaction situation with some intention. He or she also carries into the situation a body of pre- existing knowledge, in the form of memories, expectations, and beliefs, which is relevant to the goals of the interaction.
Place Figure 11.11 About Here
The actor first forms impressions of the situation and the target. In so doing, he or she combines whatever information is available in the environment -- the perceptual features of the setting and the target, for example -- with inferences derived from pre-existing knowledge. This impression then serves as a guide to planning and executing some action directed at the target. The target then interprets the meaning of the actor's behavior. This interpretation is shaped by his or her own impressions of the situation and the actor, formed on the basis of perceptual information and pre-existing memories, expectations, and beliefs, as well as by his or her goals in the current interaction setting. The target then responds to the actor -- a response that is determined by his or her own interpretation of the actor's behavior. The actor then interprets the target's response, perhaps revises his or her his or her impression of the actor, and the cycle starts all over again.
But it is not only the target who must interpret the actor's behavior, and the actor who must interpret the target's response. In situations where environmental constraints are powerful, or where the behavior has been determined by automatic or unconscious mental processes, the participants may well be surprised by their behavior. Under those circumstances, they will be forced to interpret their own behavior as well as that of the other person. Deliberate planning activity does not necessarily precede all behavior, for reasons that will be discussed later. However, it is an extremely likely to occur when behavior or its consequences surprises the person. Thus, the general social interaction cycle, consisting of the transactions between two individuals, also subsumes two intrapersonal cycles, one within the actor and one within the target.
Let's illustrate the social interaction cycle with a familiar example: Jack and Jill, a young man and a young woman, in a dating relationship. They know each other, they like each other, they have dated before, and those dates went well. Now they have come together on yet another Saturday night for the purpose of having a date. Some aspects of the date have already been planned in advance: they will go to a local Mexican restaurant and then to a club where a jazz group is playing. Some aspects of the date have yet to be negotiated: Jill would like to escalate their sexual relationship somewhat (they have kissed but nothing more), but Jack's isn't yet sure enough about his feelings toward Jill. During dinner, Jill strokes Jack's arm affectionately, and at the club she holds his hand tightly through the first set. At the break, she suggests that they repair to her apartment, and Jack hesitantly agrees.
Next we find Jack and Jill at home, on the couch, fully clothed. After some small talk Jill puts her arms around Jack, and kisses him on the lips. They have done this before, and Jack responds by holding her closer and kissing her harder and longer. Soon they are locked in passionate embrace. Jill takes this as permission to escalate: she slips her fingers inside Jack's shirt and strokes his chest. He becomes even more aroused, and tentatively reaches for the buttons on her blouse. After undoing two or three, he realizes that he is going further than he wanted to just now, and he pulls away. Jill has never experienced anything like this before -- most of they time she has had to fend off advances from her dates. Now, for the first time, she begins to doubt Jack's feelings for her. She may even doubt that she is attractive enough to interest Jack as a lover as well as a friend. She says, "I thought you liked me, I thought you liked making out". Jack replies that he does. So after an interval of silence she decides to try another tack. Another embrace, more kissing, and again Jack pulls away. Finally he says, "I do like you, Jill, but I'm just not ready for a heavy sexual relationship just yet. Can't we go a little more slowly, at least for a while?" Jill realizes that Jack's behavior has nothing to do with her, and that he may have been a little frightened at her aggressiveness. She backs off, they curl up together on the couch to watch a little television, and Jack goes home.
Here we have all the essential features of the
social interaction cycle: two people engaged in a
goal-directed encounter, but each with somewhat different
goals, different histories, and different expectations. Jill
wants a little more sex in her life, and her experience with
other dates leads her to think that Jack will comply. Jack
isn't sure he's ready for sex, and since he's heard about
the resistance his other male friends have encountered from
their dates, he figures he's safe for a while. When Jill
turns on the heat, Jack can't help reciprocating -- as, in
fact, she expected him to do. But when he realizes that he's
doing things he doesn't want to do, he stops. Until he
explains himself, his behavior is puzzling to Jill. Once she
understands the situation from his point of view, she can
understand his behavior as well. (Jack and Jill finally work
things out, but we leave the details as an exercise for the
Analysis of the Social Interaction Cycle
The story of Jack and Jill is fine as an
anecdote, but it remains to be seen whether such processes
can actually be studied in the psychological laboratory.
Obviously, researchers don't bring dating couples together
with a couch and a television set. However, social
psychologists have developed ingenious techniques for
simulating various components of the cycle in the
Consider, first, a variant on the "getting acquainted" paradigm often used by psychologists to study social interaction. In this type of study, two strangers, are brought together under controlled conditions, and the investigator studies how they develop knowledge about each other. In one study (Snyder & Swann, 1978a), the two participants were of the opposite sex. Males were assigned to the actor role, while females were assigned the target role. At the outset of the experiment, the actor was given a personality profile that described the target as relatively extraverted or relatively introverted. This information was completely unrelated to the her actual personality -- in fact, most of them, like most women (and most men, for that matter) were neither especially extraverted nor especially introverted. The actor was instructed to select some questions from a set provided by the experimenter, to be asked over the telephone in order to connect the target's behavior with her personality profile. Some of the questions were highly relevant to extraversion (e.g., "Tell me how you would liven up a dull party"), some highly relevant to introversion (e.g., "Tell me why you find it difficult to make friends"), and some were irrelevant to introversion-extraversion. Not surprisingly, the actors tended to select questions that were related to the profile they had been presented earlier. They asked questions about extraversion of women they thought were extraverted, and about introversion of women they thought were introverted.
The success of this deception depends on the variability of human social behavior. Almost everybody, no matter how extraverted, has had trouble making friends at one time or another. And almost everybody, regardless of how introverted they are, knows some strategies for enlivening a dull party. Thus, the bias in the questions permitted each of the targets to display behavior that was consistent with the profile of her personality, regardless of whether that profile was accurate. The targets' replies were recorded. A group of naive judges listened to the recordings and then rated the target's personalities on the basis of what they heard. The results are shown in Table 11.6. Targets initially labelled as extraverted were rated higher on extraversion than those labelled as introverted; conversely, targets initially labelled as introverted were rated higher on introversion than those labelled as extraverted. Apparently, questioning by the biased actor elicited responses from the target that were consistent with the actor's expectations -- responses that led even a naive judge to believe that the women were indeed introverted or extraverted.
Place Table 11.6 About Here
Similar findings were obtained in another study
employing a different manipulation of the actor's
expectations. In this case, the actors (again, men) were
presented with ostensible photographs of the targets (again,
women) rather than narrative personality profiles. The
individuals in the photographs were either fairly attractive
or fairly plain (though not unattractive) women. It is a
common stereotype (but, like all stereotypes, untrue) that
attractive people are relatively gregarious and sociable,
while introverted people are shy and withdrawn. The actors
tended to agree with these stereotypes. When asked to
describe their targets' personalities, before they ever had
any direct experience with them, they tended to describe the
attractive individuals as poised, sociable, and humorous;
they described the plain individuals as awkward, unsociable,
and serious. These anticipatory images were clearly related
to their behavior in a subsequent telephone conversation.
And, even though the targets were randomly assigned to the
attractive and plain conditions, the actors' behavior was
contagious. A group of naive judges rated the targets'
personalities on the basis of listening only to their side
of the telephone conversation. As shown in Table 11.6, the
judges rated the "attractive" targets as more extraverted
than the "plain" targets. In this experiment, as in the
other one, extraverted and introverted behavior was created
were none had existed before, solely by the actor's
The Self-Fulfilling Prophecy
Experiments such as these are demonstrations of the self-fulfilling prophecy. Merton (1948, 1957) defined the self-fulfilling prophecy as "a false definition of the situation evoking a new behavior which makes the originally false conception come true. Darley and Fazio (1980) redefined the self-fulfilling prophecy in terms of expectancy confirmation: actors behave in such a way as to evoke events that confirm their expectations, regardless of whether these expectations were appropriate in the first place. They also distinguished between two kinds of expectancy confirmation processes. In perceptual confirmation, ambiguous events are interpreted as consistent with one's expectations, although observers who do not share the expectation would not necessarily concur with the interpretation. In behavioral confirmation, the events themselves are unambiguous, so that both biased and unbiased observers would agree on their interpretation.
Although the concept of the self-fulfilling prophecy quickly became part of our everyday vocabulary, the phenomenon itself was not demonstrated experimentally until the 1960s in a line of extremely provocative research by Rosenthal and his associates. In one study, undergraduates in an experimental psychology laboratory course were assigned to train white rats to run a maze (Rosenthal & Fode, 1963). Half the students were told that their rats were from a strain of animals selectively bred to be "maze bright" -- that is, they learned mazes easily. The others were told that their rats were "maze dull". In fact, of course, the rats were as identical as laboratory animals can be. Nevertheless, the rats labelled as "maze bright" achieved faster running times than those labelled as "maze dull". The experimenters were took care to eliminate the effects of misreporting on the part of the students: so far as we can tell, the "maze- bright" animals really did perform better than their "maze-dull" counterparts. Apparently, the students' expectations led them to treat their rats -- during the pretest handling phase, perhaps -- in such a way as to facilitate or impair their maze-learning performance.
Further evidence for expectancy confirmation effects was provided by a study of academic performance (Rosenthal & Jacobson, 1968). These investigators studied children in an elementary school situated in a lower- working-class urban area. The school had the traditional six grade levels (excluding kindergarten), and in each the children had been assigned to classrooms based on their level of scholastic ability - low, medium, and high -- as determined by standardized tests. Under the guise of validating a new test designed to predict academic progress, the investigators administered a nonverbal IQ test to the children in each class at the beginning of the school year. Later, the teachers were told which of their pupils had scored as "intellectual late bloomers" -- that is, pupils who would show exceptionally good progress during the year, regardless of how they were doing at the moment. What the teachers were not told was that these children, comprising approximately 20% of each classroom, had been randomly assigned to the late-bloomer group: in terms of test scores, they were indistinguishable from the others in the class. At the end of the school year the students were re-administered the IQ test. As might be expected, the average student showed a gain in IQ score since the beginning of the year -- despite what some would believe, school does tend to make students smarter! More surprising, the students who had been labelled as late bloomers showed, on average, significantly greater gains than those who had not been so labelled. Apparently, the teachers behaved towards the late-bloomers in such a manner as to elicit from them precisely the academic gains that had been prophesied.
Rosenthal and Jacobson called their study "Pygmalion in the Classroom". Although we do not know precisely how the undergraduates shaped the behavior of their laboratory rats, later research shed considerable light on how the teachers achieved their results. Observational studies showed that their treatment of late-bloomers differed from that of control pupils in several ways. Socio-emotional climate: they tended to be more warm, friendly, encouraging, and attentive toward late- bloomers.Feedback: they gave late-bloomers more performance-contingent phrase.Input: they gave late-bloomers more, and more challenging, assignments.Output: they encouraged more classroom participation in the late-bloomers, prompting and shaping their answers in various ways. This is not to say that teachers were doing these things deliberately. For the most part, they were quite unaware of how they were treating their pupils differently. When actors are aware of the possibility of a self-fulfilling prophecy, its effects tend to be reduced.
Although Rosenthal's initial reports were
highly controversial (e.g., Barber & Silver, 1968;
Elashoff & Snow, 1971). However, literally hundreds of
studies in many different domains leave little doubt that
expectancy confirmation effects are genuine (Rosenthal,
1974, 1976; Rosenthal & Rubin, 1978). Apparently,
teachers, supervisors, therapists, experimenters, and
ordinary people on the street act in such a way as to elicit
from their interaction partners behavior that confirms -- or
is perceived to confirm -- their expectations of
Sequential Analysis of Expectancy Confirmation Processes
An experiment by Snyder and Swann (1978b) shows in a rather precise manner how the self-fulfilling prophecy transpires. In the experiment, three subjects who were unknown to each other were scheduled for a study of reaction time. The experimenters told the subjects that in order to stimulate interest (reaction-time experiments, while important for science, can be painfully boring) the session would be run as a game, with a bonus to the subject who achieved the fastest average response latencies. The subjects were told that while one subject sat at the reaction time apparatus, the other subject was given control of a "noise weapon" that he could use to disrupt the performance of his "opponent", and thus gain an advantage in the contest. After several trials the subjects would switch positions, so that the victim would get a chance to retaliate. In an elaborately choreographed procedure, two subjects were selected to participate in the first phase of the experiment, and the third was taken to a waiting room.
In Phase I, the two subjects first filled out a questionnaire concerned with hostility, competitiveness, and aggression. Then one subject, called the "target", was randomly assigned to sit at the reaction time apparatus; the other, called the "labeling perceiver", was seated at the noise weapon. Unbeknownst to the target, the labeling perceiver was given a personality sketch of his opponent. The profile, ostensibly based on the personality questionnaire, described the target as either highly hostile (for half the labeling perceivers) or as highly cooperative (for the other half). Needless to say, these descriptions had nothing to do with their target's actual personalities. At the same time, and unbeknownst to the labeling perceiver, the targets were given an attribution for their own behavior in the experiment. Half the targets were told that previous research indicated that individuals' behavior with the noise weapon was determined by the kind of person they are (a dispositional attribution), while the other half were told that it was determined by the way they were treated (a situational attribution). (Needless to say in a chapter on interactionism, behavior is not strongly determined by either dispositional or situational factors).
Under these circumstances, Snyder and Swann argued that the intensity selected by the operator of the noise weapon was an index of hostility. On the first round, with the target seated at the reaction-time apparatus and the labeling perceiver at the noise gun, ??% of the perceivers anticipating an interaction with a hostile partner selected a high intensity setting, compared to only ??% of those anticipating a cooperative one. Thus, they treated ostensibly hostile partners in a hostile manner, and ostensibly cooperative ones in a cooperative manner. On the second round, with their positions reversed, those targets who had been labelled as hostile, in fact, selected a higher intensity setting than those who had been labelled as cooperative. After three more turns each, Phase 1 was concluded. When the perceivers were asked to rate their opponents, the targets labelled as hostile were actually rated as more hostile than those labelled as cooperative. Thus, people who were inappropriately labelled as hostile, and treated as such, actually came to behave in a hostile manner.
For Phase 2 the labeling perceiver was dismissed, and the third subject was introduced. Because he had no information, accurate or otherwise, about the target's personality, he is called the naive perceiver. When the round began, the naive perceiver was seated at the reaction-time apparatus, the target at the noise gun. Again, targets labelled as hostile selected higher intensities than those labelled cooperative. But this was the case only for those who had been given a dispositional attribution for their behavior -- that is, for those who had been encouraged to believe that their behavior reflected the kind of person they were. Those given a situational attribution showed no differences in their treatment of their new partners. After three more rounds, the naive perceivers were asked to rate their opponents. Given what transpired in Phase 1, it is no surprise that targets initially labelled as hostile were continued to be rated as more hostile than those initially labelled as cooperative -- but, again, only those given a dispositional attribution for their own behavior. For those given a situational attribution, there were no rating differences.
In summary, behavior created by the
expectations of the labeling perceiver persisted with the
naive perceiver -- provided that the targets possessed
certain expectations about themselves. In this
experiment, we actually see two self-fulfilling
prophecies. One, created by the expectations of the labeling
perceiver, is fulfilled when the perceiver acts in a manner
congruent with his expectations and the target responds in
kind. The other, somewhat paradoxically, is created by
expectations that the targets form of themselves. Consider
the case of the targets labelled as hostile. They are
treated in a hostile manner, and so it is not surprising
that they retaliate, act in a hostile manner themselves,
when given the opportunity to do so. Recall, however, that
some targets have been told that their behavior is
determined by the kind of person they are, rather than the
situation into which they have been placed; the others, have
been encouraged to make precisely the opposite attribution.
"The kind of person one is" carries implications of
stability and consistency; "the situation one is in" carries
implications of variability and flexibility. Thus, the
behavior of the targets is determined by how they perceive themselves
as much as it is by how they are perceived by others. Thus,
although the self-fulfilling prophecy is generally thought
of as a pure situational effect -- we become whatever others
think we are -- the Snyder and Swann study shows that it
actually is a product of the interaction of personal and
Expectancy Confirmation and Self-Verification
In most studies of expectancy confirmation, the actor's behavior is created by an experimental manipulation -- the information provided by the experimenter, which leads the person to form specific expectations about some person, or situation, or even him- or herself. These expectations bias the perceiver's perception of current events, retrieval of memories from the past, information-seeking and hypothesis-testing, and consequently the actions that are guided by these cognitive processes. The actor's behavior elicits reciprocal actions from the target -- behaviors that either confirm the actor's expectations or are interpreted as doing so.
Given the short-term effects of experimentally induced expectations, it is only a small step to consider the long-term effects of pre-existing expectations. Suppose, for example, that a person has developed a generalized expectation that others are hostile and competitive, or a more specific expectation about some group of people (for example, stereotypes about men or women, blacks or whites, Christians or Jews, gays or straights, Germans or Russians), or an even more specific expectations about a particular person. Just as surely as in the controlled experiment, such expectations can function as self-fulfilling prophecies. These expectations, thus confirmed, can lead to subsequent exchanges that further strengthen the original expectation. Moreover, self-perception processes can lead the targets to adopt false impressions of themselves that match the false expectations of the actors. When the original expectation is relatively positive, as in the case of the Pygmalion experiment, the outcome of this process may be beneficial and desirable. When the expectation is negative, however, as in the case of most social stereotypes, the process can degenerate into a vicious cycle in which we, and those with whom we interact, fall to the mercy of our expectations.
When the actor is aware of the pernicious effects of the self- fulfilling prophecy, these effects can be eliminated to a considerable degree. In addition, repeated exposure to expectancy-disconfirming (or schema-incongruent) information can lead to changes in our expectations and consequent behavior. The problem, of course, is that we are not always aware of the effects of our expectations on others' behavior. And of course, behavioral-confirmation effects operate precisely to prevent the actor from being exposed to expectancy-disconfirming behavior --a a process that is further abetted by biases in information-seeking and hypothesis- testing. Finally, even when the target displays ambiguous or frankly disconfirmatory behavior, perceptual confirmation processes will operate to distort their interpretation. For this reason, erroneous expectations, which should be self-correcting, are not necessarily corrected unless the target exerts some degree of counter-control over the process.
As it happens, targets can take active steps to correct the actor's expectations, and thus his or her behavior. This process, known as self- presentation (Goffman, 1959) or impression-management (Jones, 1965), is the opposite side of the coin from impression-formation. At the same time as one person attempts to form an impression of another participant in a social interaction, the target of this process is trying to create a particular impression of him- or herself in the mind of the perceiver. This process of deliberately shaping others' impressions of us is sometimes known as strategic self-presentation, because its goal is often to gain power or advantage over another in some specific social interaction (and thus maximize one's own outcomes in social exchange. However, as we shall wee, self-presentation can also be genuine, in that the person's goal is to present an accurate portrayal of him- or herself, rather than to gain power.
A partial listing of impression-management strategies was provided by Jones and Pittman (1980), and are listed in Table 11.7. In ingratiation we attempt to create an impression of ourselves as likeable by overt self- characterization, conforming to the opinions of others, enhancing the status of others, and doing favors for those whom we would have like us. In intimidation we use threats of aggression, anger, and mental breakdown in order to convince others that we are dangerous. In self-promotion, claims and accounts of past performance, real or imagined, tend to convince others that we are competent (actual displays of competent behavior also help!). In exemplification, we employ such techniques as self-denial, helping others, and militant advocacy to suggest that we are worthy of other people's admiration, respect, and trust. And in supplication, we portray ourselves as helpless through pleas for help and expressions of self-deprecation.
Place Table 11.7 About Here
This is only a partial list of impression-management strategies. There are many others, and each of us has our own favored impression -- or, more likely,set of impressions -- that we wish to inculcate in the minds of other people. However, the successful impression-management depends on other people reading us the way we want to be read. In fact, studies have shown that there is a great deal of consensus about the meaning of various behaviors, and their implications for the personality of the person who displays them (Jones, 1965; Jones & Berglas, 19??; Jones & Pittman, 1980; Buss & Craik, 1980).
The fact that impression-management works as well as it does also testifies to the cultural consensus that surrounds the interpretation of behavior. In the present context, the success of strategic self- presentation means that targets can use various impression-management strategies to correct actors' erroneous expectations and impressions, and to conserve their subjective impressions of themselves. Swann (1983) has referred to this process as self-verification.
Consider first an experiment by Swann and Hill (1982), employing a variant on the noise-weapon paradigm. In this case, two people were brought together for a study of game-playing behavior. At the beginning of the session they were administered a questionnaire measure of dominance, on the basis of which they were classified as dominant or submissive. They were then asked to play a game similar to Mastermind. The experimenter constructed an array of colored pegs, and the two subjects were asked to work together to try to guess their colors and locations. One subject was actually a confederate of the experimenter, although of course this was not known to his partner.
Through an ostensibly random process, the confederate was assigned to the role of decision-maker, and the subject to the role of advice-giver. In subsequent conversation, the confederate made remarks about the subject's personality that were either consistent or inconsistent with his scores on the questionnaire administered earlier. In the consistent condition, for example, he might say to a dominant subject, "I bet you don't like this assignment -- you strike me as a dominant kind of person". In the discrepant condition, however, he might say to the same subject, "I bet the assignment is all right with you -- you strike me as a submissive kind of person". This feedback manipulation was intended to inform the subjects that their partner had an impression of them that was either consistent, or inconsistent, with their impressions of themselves.
After this, half the subjects were given an opportunity to react to the confederate's statements. Their behavior was rated by naive judges for resistance to the confederate's comments and expressions of dominance or submissiveness. Dominant subjects given discrepant feedback showed substantially more resistance, and substantially more dominant behavior, than those given consistent feedback. Similar findings were obtained with submissive subjects with respect to their submissive behavior. Not surprisingly, a post-experimental administration of the personality questionnaire revealed that these subjects showed relatively little change in their self-concepts. (Subjects given consistent feedback, of course, showed even less change.) However, subjects given discrepant feedback, but no opportunity to correct their partners' impressions of them, showed relatively large shifts in the direction of the feedback. Apparently these subjects, who were given no self-presentation opportunity, began to believe what had been said about them!
It may seem hard to believe that effects on self-perception could occur following such a relatively trivial, short-term manipulation. Certainly, it is hard to believe that a person like Mother Theresa could be led to think of herself as a misanthrope on the basis of malicious whispers in alleys of Calcutta. Expectancy confirmation effects can be counteracted, but this is most likely to occur when the subject has a stake in the expectancy -- that is, when the prophecy conflicts with some firmly established portion of the target's self-concept. Under such circumstances, we are faced with a kind of "battle of will" (Swann & Ely, 1984, p. 1287) between perceptual and behavior confirmation processes on the part of the perceiver, and self-verification processes on the part of the target. In such a battle, which side will win?
The question was addressed in another experiment by Swann and Ely (1984), employing a variant on the getting-acquainted paradigm. Two individuals, previously unacquainted, were brought together. One of the pair, the perceiver, received some information about their partner's personality. They were told that the target was either extraverted or introverted; further, the experimenters manipulated the degree of certainty attached to the information. The information provided to the perceiver was always discrepant with the target's personality, as assessed by a pre- experimental questionnaire. However, the targets were selected for the degree of certainty attached to their self-description. For example, subjects who scored extremely high on the scale of extraversion were assumed to be more certain of their position on this dimension than those who achieved more moderate scores.
In Phase 1, the perceivers asked questions of the targets about their personalities. As might be expected, they tended to probe for evidence that would confirm their expectancies. And, given the degree of flexibility shown by even the most introverted or extraverted subjects, they often got what they were looking for -- but not always. When naive judges rated the target's replies, the behavior of the targets who were uncertain about themselves tended to conform to the perceiver's expectations; this was especially the case when the perceiver was highly certain. By contrast, the targets who were sure of themselves behaved in accordance with their self-description, regardless of the the certainty of the perceiver.
In Phase 2 of the experiment, the perceivers were given another opportunity to confirm their expectations of the self-sure targets. Having been rebuffed once by the self-sure targets, they might be expected to persist with the self-fulfilling prophecy. On the contrary, they shifted their approach, now tending to probe for information that was inconsistent with their initial expectations. This was especially the case when they were uncertain of the targets personalities. In Phase 3, their questions were unrelated to their expectations. In both phases, the judges rated the targets as behaving even more strongly in accordance with their self- impressions than they had previously -- especially when they were certain about their own personalities. Thus, the perceivers began by soliciting behavior that was consistent with their expectations, but the targets displayed behavior that was consistent with their impressions of themselves. In other words, the "battle of wills" is resolved in favor of the target. Provided that they are given an opportunity to engage in impression management, and have a clear idea of the impression they want to manage, people can counteract the self-fulfilling prophecy, and express their own personalities.
Although this would seem like a case of the
determination of behavior by personal rather than
situational factors, it is important to bear two matters in
mind. First, for many targets, behavior was determined
largely or wholly by the expectations of the perceiver --
after all, with respect to introversion-extraversion, they
possessed no internal determinants to counteract the effects
of the situation. Second, and more important, the personal
determinants important to the behavior of other targets were
themselves the product of the complex interplay of personal
and situational factors. In this way, behavior always
reflects some mix of the personal and the situational, and
always feeds back to alter precisely those factors that
determined it in the first place.
Toward a Cognitive Psychology of Personality
The analysis of the social-interaction cycle in
terms of expectancy confirmation and self-verification
processes leads to the same conclusion as our earlier
analysis of the person-environment interaction and
reciprocal inhibition. The features of the person that play
the most important role in guiding social interaction are
cognitive in nature. What people think of themselves, their
interaction partners, and the situation in which interaction
takes place seems much more important than anything else
about them. Personality reveals itself in social
interaction, and social interactions, whether mundane or
monumental, are intelligent. When normal and adaptive, our
relations with each other are discriminating, not
indiscriminate; flexible, not rigidly stereotyped, and
optional, not obligatory. These are the hallmarks of
intelligent behavior, and they are best understood in terms
of human intelligence -- in the ability of people to think
about themselves and what they are doing, and to control
their behavior in light of the current situation, past
experience, and the anticipated future. Theories of
personality built around non-cognitive behavioral
dispositions, or around primitive motivations are not
enough. In addition, perhaps instead, we need a theory of
individual differences between that takes as its centerpiece
the intelligence and cognitive capacity that are the
hallmarks of human nature.