ATTITUDE TOWARD OBJECT AND ATTITUDE TOWARD SITUATION AS PREDICTORS OF BEHAVIOR - Thesis for the Degree of M; A. * I MICHIGAN STATE UNIVERSITY PETER T. KLIEIUNAS 1969 v-th‘v‘o— dQ'W-«uco-uw. . V ‘.. K. -. . r; ABSTRACT ATTITUDE TOWARD OBJECT AND ATTITUDE TOWARD SITUATION AS PREDICTORS OF BEHAVIOR BY Peter T. Kliejunas The concept of attitude and the relationship between attitudes and behavior are major concerns of both social psychology and personality theory. Relatively few studies, however, have systematically examined this relationship. Those few studies which have tested the assumption that a person's behavior will be consistent with his attitudes have generally found a lack of correspondence between ver- bally expressed attitudes and actual behavior. The relationship between attitudes and behavior is examined here in light of Rokeach's (1968) recent definition of the nature of attitudes. Rokeach has postulated that a person's social behavior must always be mediated by at least two types of attitude: an attitude activated by an object (A0), and an attitude activated by the situation in which the object is encountered (As). Behavior (B) is a function of AOAS, a configuration produced by the cognitive interaction between A0 and As. The outcome of this inter- action, and thus behavior, will depend upon the relative Peter T. Kliejunas importance of A0 and AS with respect to one another in the context AOAS. Eighty-one university students enrolled in an intermediate psychology course were used as subjects to determine correlations between B, reported number of cuts in psychology courses, and three attitudinal measures: (a) A0, evaluations of course instructors, (b) AS, evalu- ations of general importance of attending classes, and (c) AoAs’ a weighted average of A0 and AS derived from an adaptation of Rokeach and Rothman's (1965) belief con- gruence model of cognitive interaction. Supporting all of the hypotheses, results show that: (a) AOAS is a better predictor of B (p < .001) than is either A0 or AS, (b) A0 and AS are uncorrelated, (c) AOAS is as good a predictor, or better, of B than is the multiple correlation using B as the criterion variable and A0 and AS as the predictor variables, (d) AS is generally a better predictor of B than is A0, and (e) the mean evalu- ation of the situation, and of the perceived importance of the situation, is greater than the mean evaluation of, and the perceived importance of, the objects encountered in that situation. The results suggest that: (a) attitudes, properly conceptualized and measured, can be accurate predictors of behavior, and (b) the importance of situational attitudes and their interactions with attitudes toward objects have Peter T. Kliejunas been generally overlooked in past research on the relation- ships between attitudes and behavior. , A ') _ Approved: )Zngrh LL/g/Lgig/Z Date: 74;. /' 7 /7"70 ATTITUDE TOWARD OBJECT AND ATTITUDE TOWARD SITUATION AS PREDICTORS OF BEHAVIOR BY Peter T. Kliejunas A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Psychology 1969 Q (Hz/xii a/. «4/- 70 ACKNOWLEDGMENTS I wish to gratefully acknowledge and to thank all those who have aided in the completion of this thesis. I especially wish to express my sincere appreci— ation to my advisor and committee chairman, Dr. Milton Rokeach, whose insights and suggestions have formed the basis of this study. The always generous contributions of his time, in which he has offered valuable guidance, shown considerable patience, and given continuing encouragement, have made the planning and the execution of this research a most meaningful and rewarding learning experience. I am also indebted to the other members of my committee, Dr. Charles Hanley and Dr. William Crano. Giving freely of their time, they have offered many valuable com- ments and suggestions which have been of distinct aid in the completion of this thesis. Thanks are also due to Dr. Mark Rilling, who generously offered to give up a portion of his class time on two different occasions so that the data upon which this research is based might be collected. ii TABLE OF CONTENTS Page ACKNOWLEDGMENTS O O O O O O O O O O O O i i LIST OF TABLES O O O O O O O O O O O 0 iv INTRODUCTION 0 O O O O I O O O O O O O 1 Centrality of the Concept of Attitude. . . . 2 Attitudes and Behavior. . . . . . . . 3 Previous Studies of Attitudes and Behavior . . 5 Situational Variables and Behavior. . . . . 9 The Need for a Reconceptualization of the Relationship Between Attitudes and Behavior. 11 Attitude Toward Object and Attitude Toward Situation 0 O O O O O O O I O O O i 12 Cognitive Interaction Between A0 and AS . . . l9 Hypotheses. . . . . . . . . . . . . 22 METHOD 0 O O O O O O O O O O O O O O 2 6 Subjects and Procedure. . . . . . . . . 26 The Questionnaire . . . . . . . . . . 27 Analysis of the Data . . . . . . . . . 31 RESULTS 0 O O O O O O O O 0 O O O O O 3 5 subsamp 1e ET 0 O O O O C O O O O O O 3 5 SUbsamp 1e B1 0 o o o o o o o o o o o 4 l Subsamples B , B , and B . . . . . 46 Summary of Résulés i3 ReIation t8 the ‘ Hypotheses. . . . . . . . . . . . 65 DISCUSSION. . . . . . . . . . . . . . 71 The Conceptualization of Attitude . . . . . 72 Some Implications for Previous and for Future Research . . . . . . . . . . . 75 Limitations of Present Study. . . . . . . 80 REFERENCES 0 O O O O O O O O O O O 0 O Q 8 7 iii Table 10. ll. 12. LIST OF TABLES Ranges, Means, Variances, and Standard Deviations of Obtained Measures for Subsample BT . . . . . . . . Intercorrelations Between the Obtained Measures for Subsample BT. . . . Ranges, Means, Variances, and Standard Deviations of Obtained Measures for Subsample Bl . . . . . . . . Intercorrelations Between the Obtained Measures for Subsample Bl' . . . Ranges, Means, Variances, and Standard Deviations of Obtained Measures for Subsample B2 . . . . . . . . Intercorrelations Between the Obtained Measures for Subsample B2. . . . Ranges, Means, Variances, and Standard Deviations of Obtained Measures for Subsample 83 . . . . . . . . Intercorrelations Between the Obtained Measures for Subsample B3. . . . Ranges, Means, Variances, and Standard Deviations of Obtained Measures for Subsample B4 . . . . . . . . Intercorrelations Between the Obtained Measures for Subsample B4. . . . Ranges, Means, Variances, and Standard Deviations of Obtained Measures for Subsample B5 . . . . . . . . Intercorrelations Between the Obtained Measures for Subsample BS' . . . iv Page 36 39 42 44 47 49 51 53 54 56 58 6O Table Page 13. Ranges, Means, Variances, and Standard Deviations of Obtained Measures for Subsample B6 . . . . . . . . . . 61 14. Intercorrelations Between the Obtained Measures for Subsample B6' . . . . . 63 15. Summary of Results Pertaining to A0, A5, and AOAS as Predictors of B. . . . . . . 66 16. Summary of Results Pertaining to Mean Evaluation and Mean Perceived Importance of Situation and Objects . . . . . . 69 INTRODUCTION The concept of attitude is generally regarded as central to both social psychology and to personality theory. In spite of, or perhaps because of, its prominent position, there presently exists much confusion and vague- ness regarding the definition and the referents of the concept. If there is any common core of agreement evident in the proliferous literature regarding the concept of attitude, it is that attitudes are in some way related to behavior. More often than not, this relationship is merely as assumption or an inference. There are surprisingly few studies which have systematically examined the relationship between attitudes and behavior. Those studies which have attempted to study this relationship have generally found a lack of corres- pondence between overt behavior and verbally expressed attitudes. This lack of correspondence is often attributed to inadequacies of definition or of measurement or is ex- plained away by some nebulous allusions to the influences of situational variables. It has also led to conclusions that behavior cannot be predicted on the basis of attitude test scores alone. The purpose of the present study is to examine the relationship between attitudes and behavior in the light of Rokeach's (1968) recent definition of the concept. It will be shown that when this conceptualization of attitude is employed, and when appropriate measurements are made, a person's behavior can indeed be predicted from knowledge of his attitudes. Before turning to Rokeach's definition and to the hypotheses and procedures of the present study, it will be beneficial to discuss some previous conceptuali- zations of attitude and the relationships between attitudes and behavior. Some of the studies which have addressed themselves to this relationship will also be discussed. Centrality_g£ the Concept of Attitude Reviewing the history of attitudes in social psychology, G. W. Allport in 1935 referred to the concept as "probably the most distinctive and indispensible con— cept in contemporary social psychology. No term appears more frequently in experimental and theoretical literature" (p. 798). In 1937, Murphy, Murphy, and Newcomb emphasized the same point: "Perhaps no single concept within the whole realm of social psychology occupies a more nearly central position than that of attitudes" (p. 889). Allport (1935) assigns credit for the establishment of the con- cept of attitude as a permanent and central feature of social psychology to Thomas and Znaniecki. The latter authors, in their study of the Polish peasant (1918), defined social psychology as the scientific study of atti- tudes. Other early writers have also equated social psychology with the study of attitudes (e.g., Bogardus, 1931; Folsom, 1931). Social psychology and sociology, however, are not the only disciplines in which the concept of attitude is central. Books such as Escape from Freedom (Fromm, 1941) and The Authoritarian Personality (Adorno, Frenkel-Brunswik, Levinson, and Sanford, 1950), have established the rele- vance of attitudes for personality theory. The concept of attitude is also becoming increasingly popular among politi- cal scientists (see e.g., Campbell, Converse, Miller, and Stokes, 1960, and Dreyer and Rosenbaum, 1966). The concept of attitude and its referents have been much refined and extended over the years. It is still, however, regarded as central. Fishbein, writing in 1967, states that . . . despite the enormous growth of social psychology, and the diversity of interest of contemporary social psychologists, Allport's words are as true today as they were in 1935. In addition, the attitude concept has come to play an increasingly important part in almost all of the behavioral sciences and many of the applied disciplines (p. v). Attitudes and Behavior The major reason for the prominent position of the concept of attitude, both in the past and today, seems clear. Attitudes are viewed as being related in some way to behavior. Most often, this relationship is viewed as a causal one. An individual's behavior depends upon or is in some way influenced by his attitudes.l If nothing else, there should at least be consistency between an individual's attitudes and his behavior. The concern of the behavioral sciences is, by definition, behavior. If our task is to explain and/or to predict behavior, then we must study attitudes. This relationship, sometimes implied and some- times explicit, is brought out by the following representa— tive definitions and characterizations of attitudes. The attitude, or preparation in advance of the actual response, constitutes an important determinant of the ensuing social behavior (F. H. Allport, 1924, p. 320). [Attitudes are] acquired tendencies to act in specific ways toward objects (Krueger and Reckless, 1931, p. 238). An attitude is a mental and neural state of readiness, organized through experience, exerting a directive or dynamic influence upon the individual's response to all objects and situations with which it is related (G. W. Allport, 1935, p. 810). An attitude, whatever else it may be, denotes a functional state of readiness which determines the organism to react in a characteristic way to certain stimuli or stimulus situations (Sherif and Cantril, 1946, p. 17). Attitudes are learned predispositions to respond to an object or class of objects in a favorable or un- favorable way (Fishbein, 1965, p. 107). 1The direction of the relationship may, of course, be reversed. Studies by Janis and King (1954), King and Janis (1956), and by Scott (1957; 1959), among others, have'shown the influence of behavior on attitudes. The theory of cognitive dissonance (Festinger, 1957; Brehm and Cohen, 1962) focuses primarily on changes in opinion and attitude induced by behavioral factors. It is generally assumed then that knowledge of a person's attitudes should serve as an aid in the prediction of his behavior. The relatively few studies which have attempted to predict overt behavior from knowledge of verbally expressed attitudes, however, have found little or no relationship. After more than seventy-five years of attitude research, there is still little, if any, consistent evidence supporting the hypothesis that knowledge of an indi- vidual's attitude toward some object will allow one to predict the way he will behave with respect to the object (Fishbein, 1967, p. 477). Because of this lack of evidence, some authors have questioned the basic assumption that there is a strong relationship between attitude and behavior (e.g., Fishbein, 1967). Others have tended to blame the measuring instru- ments (e.g., Cook and Sellitz, 1964), the definition of attitude (e.g., DeFleur and Westie, 1963), or both (e.g., Katz and Stotland, 1959). The present study will attempt to show that this basic hypothesis of a strong relationship between attitudes and behavior does have validity. Before focusing on the hypothesis and the procedures of the present study, how- ever, it is necessary to review some of the past research on this question. Previous Studies g£.Attitudes and Behavior One of the earliest studies to examine the relation- ship between behavior and expressed attitudes is the classical study by LaPiere (1934). Traveling extensively throughout the United States with a young Chinese couple, LaPiere had occasion to stop at many motels, hotels, and restaurants. In ten thousand miles of motor travel and in more than 250 instances, they were refused service only once. In a follow-up study, LaPiere mailed questionnaires to the proprietors of the establishments they had visited. The questionnaires asked, among other things, "Will you accept members of the Chinese race as guests in your establishment?fi Approximately 92% answered "No." There was one "Yes" reply, and the remainder answered "Uncer- tain; depends upon circumstances." Almost identical re- sults were obtained from a control group of comparable prOprietors of establishments which LaPiere and his com- panions had not visited. This early study is often cited as a striking example of the discrepancy between expressed attitudes and actual behavior. From the results of his study, LaPiere concluded that If social attitudes are to be conceptualized as partially integrated habit sets which will become operative under specific circumstances and lead to a particular pattern of adjustment they must, in the main, be derived from a study of humans behaving in actual social situations. They must not be imputed on the basis of questionnaire data (1934, p. 237). Another often cited study is that of Kutner, Wilkens, and Yarrow (1952). Three young women, two white and one Negro, entered eleven different restaurants in a fashionable northeastern community. In every restaurant visited, they were served in the same way as other patrons. Two weeks later, a letter asking for reservations was sent to the managers of these same restaurants. The letter in- cluded the sentence, "Since some of them are colored, I wondered whether you would object to their coming?". No replies to the letters were received within seventeen days. Phone calls requesting reservations for a group of persons, including some Negroes, were then made to each of the restaurants. Eight of the eleven managers said that they had not received the letters. One day later, a control phone call requesting reservations for a party of friends was made. In the control call there was no mention of race, and the reservations were accepted by ten of the eleven restaurants. As in LaPiere's (1934) study, there was apparently a large discrepancy between expressed attitudes and actual behavior. DeFleur and Westie (1958) did find a statistically significant relationship between attitudes and behavior. A scale measuring attitudes toward Negroes was adminis- tered to a group of white college students. Those who scored in the top quartile (prejudiced) and those who scored in the bottom quartile (unprejudiced) were later asked to participate in another experiment. In this second phase of the study, the recalled subjects were given projective tests which consisted of a number of photographic slides depicting pairs of Negro and white men and women in various social situations. After the administration of the projective tests each subject was told that another set of slides was needed for further re- search, and was asked whether he would pose with a Negro person of the opposite sex. The subject was then given a "standard photograph release agreement" which consisted of a graded series of situations in which the photograph might be used. The subject was asked to sign his name to each use, ranging from laboratory experiments only to a nationwide publicity campaign for racial integration, that he would authorize. The relationship between the amount of preju- dice expressed on the questionnaire and the level of signed agreement to be photographed with a Negro was found to be statistically significant. The relationship, however, was not linear. Moreover, nine of the unprejudiced and five of the prejudiced subjects (about 30% of the sample) showed inconsistency between their expressed attitudes and their behavior. DeFleur and Westie consider this too large a proportion to attribute to measurement error. They there- fore suggest the need to study intervening variables such as the social involvement of the subjects. Their general conclusion was that a one-to-one correspondence between verbal attitudes and actual behavior should not be expected. Other studies have also shown a discrepancy be- tween verbal attitudes and overt behavior. Fendrich (1967) cites a study by Saegner and Gilbert (1950) who tested the hypothesis that anti-Negro prejudice in white department store customers would not lead to discrimination against Negro sales personnel or the stores employing them. They found that there was no tendency for prejudiced individuals to avoid dealing with Negro clerks. Fishman (1961) reports that many white people with negative attitudes toward Negroes nevertheless remained in an integrated community which was becoming increasingly more Negro. Other resi- dents, however, who had positive attitudes toward Negroes, moved away. Linn (1965), in a study similar to DeFleur and Westie's (1958), also measured the relationship be- tween racial attitudes and overt behavior. He found discrepancies between verbal attitudes and subsequent overt behavior involving those attitudes in 59% of the cases. Situational Variables and Behavior In each of the investigations of the relationship between attitudes and behavior cited above, the authors mention intervening variables as possible explanations for the discrepancies which were found. Situational variables are those most often pointed to as factors other than atti- tudes which influence behavior. There are numerous examples of this influence. Lohman and Reitzes (1954), for example, studied white residents of an urban neighborhood. In the neighbor- hood situation, the subjects all belonged to a prOperty owners' association which was actively resisting Negro penetration. These same subjects, however, also belonged to a labor union which had a clear and well-implemented 10 policy of granting Negroes complete equality on the job. Biesanz and Smith (1951) report that the Canal Zone side of a street in Panama is racially segregated while the Panamanian side is racially integrated. Most Americans and Panamanians go from one side of the street to the other without any apparent difficulty. The attitudes and behavior of white coal miners toward Negroes within the same mine and outside it have been examined by Minard (1952). While racial integration and equality existed in the work situation of the mine, there was almost complete segregation above the ground. Another example is Pettigrew (1961) who emphasizes the importance of situational vari- ables in understanding the adjustments of white Southerners to racial integration. Numerous other writers have also emphasized the importance of situational variables in influencing behavior (e.g., Gordon, 1952; Lewin, 1951; Orne, 1962; Rotter, 1955). Because of the recognized influence of situational vari- ables, some authors have concluded that it is character— istics of the situation, rather than attitudes, which determine behavior toward an attitude object (e.g., Blumer, 1958; DeFleur and Westie, 1963; Raab and Lipset, 1962; Rose, 1956). Weissberg, for example, states that "an attitude no matter how conceived, is simply one of the terms in the complex regression equation we use to predict behavior" (1965, p, 427). He urges the examination of sources of influence other than attitudes on overt ll behavior, particularly the nature of the environment. Thus, directly or indirectly, several writers have suggested that little or no direct relationship between verbally expressed attitudes and overt behavior should be expected. The Need for a Reconceptualization o£_the Relationship Between Attitudes and Behavior Rather than abandoning the basic assumption of a strong relationship between attitudes and behavior, how- ever, others have focused on the need to reconceptualize this relationship. Insko (1967), for example, states that \ although "the connection between attitudes and behavior may not be a very direct one, there certainly should be some connection and it is time to thoroughly investigate this whole problem" (p. 346). Jahoda and Warren (1966) make the same point in their discussion of the few studies which have investigated the relationship between verbally expressed attitudes and overt behavior: [The lack of evidence on the relationship between attitudes and behavior] does not mean, however, that social scientists should renounce the study of atti- tudes. What it does imply is a need for theoretical models which do justice to the complexities of the relationships between attitudes and behavior, and for theory-guided investigations of these relation- ships (p. 211). Some writers have questioned the usefulness of the concept of attitude itself. It is viewed as having no systematic status as a scientific construct (Doob, 1947), or as being vague and lacking an empirical reference and thus useless in analyzing social action (Blumer, 1955). 12 Rokeach, however, feels that "the confused status of the concept can best be corrected not by abandoning it, but by subjecting it to continued critical analysis with the aim of giving it a more precise conceptual and operational meaning" (1968, p. 450). In his discussion of the nature of attitudes, Rokeach (1968) rejects the idea that there is no strong relationship between attitudes and behavior. In addition, the definition of attitude which he proposes offers a possible explanation to account for the discrepancies be- tween verbally expressed attitudes and overt behavior which have been reported in the literature. It is to this definition that we now turn. Attitude Toward Object and Attitude Toward Situation Rokeach's definition of attitude has five com- ponents: "An attitude is [l] a relatively enduring [2] organization of beliefs [3] around an object or situ- ation [4] predisposing one to respond [5] in some prefer- ential manner" (1968, p. 450). The first component of the definition points out the difference between attitudes and such momentary pre- dispositions as set or Einstellung. Attitudes are also conceived of as being organizations of beliefs. Following most theorists, an attitude is not viewed as a basic or an irreducible element within the personality. Rather, it "represents a cluster or syndrome of two or more l3 interrelated elements. In the above definition, these elements are beliefs (or cognitions, or expectancies, or hypotheses)" (1968, p» 450). Each of the beliefs within an attitude organization has a cognitive, an affective, and a behavioral component.2 An attitude, therefore, is a set of interrelated predispositions to respond. This aspect of the definition differs from other writers (e.g., Chein, 1948; Smith, Bruner, and White, 1956) who state that attitudes may or may not have a behavioral component. A response is under- stood by Rokeach to "be either a verbal expression of an opinion, or some form of non-verbal behavior" (1968, p. 453). The response itself may be affective or evaluative or may contain elements of both. It is the third component of Rokeach's (1968) definition which is most directly relevant to a possible resolution of the discrepancies between verbally expressed attitudes and overt behavior which have been reported in the literature: an attitude is organized around an object or a situation. An attitude object may be concrete or abstract, that is, a person, a group, an institution, or an issue. An individual may also, however, have a set of interrelated beliefs about how to behave in a specific 2Harding, Kutner, Proshansky, and Chein (1954) have pointed out that when ranking individuals with respect to their attitudes toward specific ethnic groups it makes little difference which of these components are used since the relationship among them is so close. l4 situation, that is, an ongoing activity or event. This second type of attitude is referred to as attitude toward situation. Rokeach (1968) points out that although attitudes have typically been defined as predispositions toward situations as well as toward objects,3 investigators have most often focused on the measurement of attitudes toward objects, across situations, rather than on attitudes to- ward situations, across objects. There are numerous scales which measure attitudes toward objects such as the Negro, liberalism-conservatism, and so on. Comparable scales for the measurement of attitudes toward situations such as managing a restaurant or buying or selling real estate, however, are very few. As a result, the study of attitudes-toward—situations has become more or less split off from the study of attitudes-toward-objects. And to account for the characteristic ways people behave with respect to specific social situations, altogether new concepts are introduced, . . . trait concepts . . . role con- cepts . . . group norm . . . definition—of-the- situation and social structure (Rokeach, 1968, p. 452). This disassociation between attitudes toward objects and attitudes toward situations has, in Rokeach's view, "resulted in a failure to appreciate that an atti- tude object is always encountered in some situation, about which we also have an organized attitude" (1968, p. 452). 3See, for example, the definitions of G. W. Allport (1935) and of Sherif and Cantril (1947) cited earlier. 15 A particular attitude object may be conceptualized as the figure, and the situation in which it is encountered as the ground. The behavior of a person with respect to an object within a situation will therefore depend on at least two factors: ". . . [1] on the particular beliefs or predispositions activated by the attitude object, and . . . [2] on the beliefs or predispositions activated by the situation" (1968, p. 455). Rokeach thus postulates "that a person's social behavior must always be mediated by at least two types of attitudes--one activated by the object, the other activated by the situation" (1968, p. 455). Kurt Lewin's well-known formula (1939) states that, in general terms, behavior (B) is a function (f) of the person (P) in interaction with his environment (E), B = f(P,E) [1] This formulation thus takes into account both the person and his environment. It must be remembered, how— ever, that Lewin is referring to the psychological en- vironment, that is, the environment as perceived by the individual. His "environment" is therefore similar to the sociological "definition-of-the-situation" (Shibutani, 1961). Thus the "environment" is not conceptually distinct from the person, and P and E cannot be separately analyzed. Many writers, however, have unwittingly moved back and forth between these two meanings of "environment." To l6 avoid the confusion involved, Rokeach prefers to use two admittedly psychological variables—-attitude toward ob- ject (A0) and attitude toward situation (AS). Lewin's formula would then be restated as follows: an individual's behavior (B) is a function (f) of his attitude toward an object (A0) and of his attitude toward the situation in which the object is encountered (AS), B = f(AO,As) [2] The proposition that behavior must always be mediated by at least two types of attitudes has an impor- tant implication. Whenever an attitude toward a particular object is activated, it will not necessarily be manifested or expressed in behavior to the same degree. The expres- sion of the attitude activated by the object will vary as the attitude toward the situation in which the object is encountered varies. The reverse, of course, is also true. Expression of behavior in a particular situation will vary as the attitude toward the object(s) associated with the situation varies. Rokeach thus states that "if one focuses only on attitude-toward-object one is bound to observe some inconsistency between attitude and behavior, or, at least, a lack of dependence of behavior on attitude" (1968, p. 455). LaPiere (1934) in the study cited earlier, for example, measured an attitude toward object (Chinese) but 17 did not take into account the different situations, and attitudes toward these situations, in which the object was encountered. The motel, hotel, and restaurant owners had attitudes not only toward Chinese, but also toward managing their establishment in a proper manner. In the face-to- face situation, they perhaps did not want to cause a scene, or embarrass LaPiere, or offend the Chinese couple whom LaPiere describes as "skillful smilers" and who spoke in unaccented English. In the non-face-to-face situation, however, the managers' attitudes toward Chinese may have been more influential than their attitudes toward the proper way to conduct a business. The discrepancies ob- served by Kutner 33 a1. may be accounted for in an analogous manner. The other studies cited above which have reported discrepancies between attitudes and behavior may be viewed in the same way. Investigators have generally focused only on attitudes toward objects and have generally ignored the equally relevant attitudes toward situations. In Minard's (1952) study, for example, the behavior of the white coal miners toward Negroes might have been expected to be different in the work situation of the mine where the safety of all depends upon a close working relationship, than outside the mine where the different situation activates different predispositions to respond. When discrepancies between verbally expressed attitudes and overt behavior are found, variables other 18 than attitudes are invoked to explain these discrepancies. As noted above, these variables are most often some type of "situational condition" such as "social involvement" (DeFleur and Westie, 1958), "commitment" (Fendrich, 1967), "the nature of the environment" (Weissberg, 1965), or "situational thresholds" (Campbell, 1963). Rokeach, how- ever, as mentioned earlier, feels that "a 'situational condition' can psychologically be reformulated as 'attitude- toward—situation' and assessed by methods similar to those employed in assessing attitude-toward-object" (1968, p. 455). Rokeach would thus disagree with writers such as Krech, Crutchfield, and Ballachey who have said that "attitude test scores alone are usually not enough to predict behavior" (1962, p. 163). Rokeach recognizes that it is not sufficient to merely assert that behavior is a function of two attitudes. It is also necessary to recognize that attitude . . . toward object and toward situation will cogni- tively interact with one another, and will have differing degrees of importance with respect to one another, thereby resulting in behavior which will be differentially influenced by the two sets of atti- tudes. In one case, an attitude object may activate relatively more powerful beliefs than those activated by the situation, thereby accounting for the generality of behavior with respect to an attitude object; or the situation may activate the more powerful beliefs, thereby accounting for the specificity of behavior with respect to an attitude object (1968, p. 456). In order to accurately predict behavior, therefore, not only must the two types of attitude be taken into account, but the cognitive interaction between attitude 19 toward object (A0) and attitude toward situation (AS) must also be assessed. This requires a theory or model of cognitive interaction. Cognitive Interaction Between A and A —o ——— —s Such a model has been proposed by Rokeach and Rothman (1965). Their belief congruence model has been shown to be approximately three times as accurate as Osgood and Tannenbaum's (1955) congruity model in quanti- tatively predicting the outcome of cognitive interaction.4 The congruity model of Osgood and Tannenbaum is an addi- tive one: the point of equilibrium is predicted on the basis of the evaluations of two objects of judgment con- sidered separately. In contrast, Rokeach and Rothman's belief congruence model is a configurationist one: the unique Gestalt formed by two components cannot be pre- dicted solely from information regarding their separate evaluations. In the belief congruence model, two elements are combined into a unique configuration which cognitively represents a . . . characterized subject (CS)--a person, thing, or idea Characterized in some unique way. The unique configuration consists of two components: a subject (S), capable of being characterized in many ways, and a characterization (C), capable of being applied to many subjects Tdeeach and Rothman, 1965, p. 129). 4Cognitive interaction is defined by Rokeach and Rothman as "the process by means of which a single evalu- ative meaning emerges as a result of combining two stimuli, each having their separate meanings" (1965, p. 129). 20 When a person is presented with a CS, and when C and S are judged to be relevant for one another, the person will compare the two components for relative importance with respect to one another. The evaluation of the CS configuration is postulated by Rokeach and Rothman to be . . . a simple average of the evaluations of C and 8 considered separately, weighted by the perceived importance of C and S relative to one another within the context CS: dCS = (w)dC + (l-w)dS [3] where dC , dc, and d5 refer, respectively, to the degree 0% polarization (positive or negative) of the characterized subject, the characterization, and the subject, and where (w) and (l-w) refer to the per— ceived importance of dc and d3 relative to one another in the context CS (1965, p. 130). In a test of the belief congruence model of cogni- tive interaction, Rokeach and Rothman (1965) had subjects first rate individual concepts and then combinations on semantic differential scales representing evaluation. Examples of component concepts are A NEGRO, RUSSIA, ANTI— COMMUNIST, and FREEDOM of the PRESS. Examples of combi- nations are A NEGRO (S) who is an ANTICOMMUNIST (C), and RUSSIA (S) extends FREEDOM of the PRESS (C). Measures of relative importance of C and s in the context of CS were also obtained in order to make specific predictions re- 5 garding the evaluation of CS. As already noted, 5These measures were obtained from a different sample. The samples were, however, considered comparable. 21 predictions made from the belief congruence model were found to be approximately three times as accurate as pre- dictions made from Osgood and Tannenbaum's congruity model. Applying the belief congruence model to the hy- pothesis that behavior is always mediated by at least two attitudes, attitude toward object (A0) and attitude toward the situation in which the object is encountered (AS), the following statements may be made. When a person encounters an object in a situation, and when A0 and AS are judged to be relevant for one another, the person will compare the two components for relative importance with respect to one another. The evaluation of the AOAS configuration or combination is postulated to be an average of the evaluations of A0 and As considered separately, weighted by the perceived im- portance of A0 and AS relative to one another within the context of AOAS. That is, 6 A A = (w)AO + (l-w)AS [4] 08 where AOAS, A0, and As refer, respectively, to attitude toward object in interaction with attitude toward situ- ation, attitude toward object, and attitude toward situ- ation, and where (w) and (l—w) refer to the perceived importance of A0 and As relative to one another in the 6Deviation (d) scores could, of course, also be used. 22 context AOAS. Since behavior (B) is postulated to be a function of A0 and As (see formula 2), formula 4 may be rewritten as B = AOAS = (w)AO + (1-w)AS [5] The purpose of the present study is to test the validity and the implications of the above postulate repre- sented by the adaptation of Rokeach and Rothman's (1965) belief congruence model of cognitive interaction. Hypotheses This postulate was tested by applying it to the case of college students cutting class. It was felt that the most relevant attitude toward object in this case would be the student's attitude toward the instructor teaching a particular course, and that the most relevant attitude toward situation would be the student's attitude toward the importance of attending class in general. Behavior (B), therefore, is the number of times a student cut a particular class, attitude toward object (A0) is his evaluation of the instructor teaching that class, and attitude toward situation (AS) is his evalu- ation of the importanceof attending class in general. Evaluations of the perceived importance of A0 and As rela- tive to one another in the context AOAS are also required to make specific predictions. The major hypothesis is, 23 then, that B will be more highly correlated with the AOAs configuration than with either A0 or AS separately. In order to adequately test this hypothesis it is necessary to show that the A0 and AS measures employed are in fact measures of different things. That is, A0 and AS must be uncorrelated. Moreover, the correlation between B and the AOAS configuration should be as high, or higher, than the multiple correlation using B as the criterion variable and A0 and AS separately as the predictor vari- ables.7 Since the sample consisted of college students, who presumably have a stake in attending classes, it was further hypothesized that the attitude toward the situation (attend- ing class in general) would be more important than the attitude toward the particular object encountered in that situation (the instructor teaching a particular course) in determining the student's behavior (cutting a particular class). In other words, although the correlation between B and the AOAS combination was expected to be higher than the correlation between either B and A0 or between B and AS, it was also expected that the correlation between B and AS would be higher than that between B and AD. This further implies that the average or mean evaluation of the situation will be higher than the mean 7Multiple correlation is, in effect, an additive model. The belief congruence model and its adaptation used here, however, are configurationist models. 24 evaluation of the objects (instructors). The mean per- ceived importance of AS in the context of AOAS, (l~w), should also be higher than the mean perceived importance of A0 in the context of AOAS, (w). In summary, the following hypotheses were tested: 1. The prediction of behavior will be more accurate when the interaction between attitude toward object and attitude toward situation is taken into account than when only attitude toward object or toward situation is con- 81dered, that 15, EB,A A > EQWA or EB,A o s o 8 Measures of attitude toward object and of attitude toward the situation in which the object is encountered will be independent of one another, that is, EAo'As ~ 0 The prediction of behavior will be as accurate, or more accurate, when the adaptation of the belief congruence configurationist model is employed than when the additive model of multiple correlation is used, that is, r >R —-B,AOAS —B AO,AS The following three hypotheses are specific to the present study. Hypotheses as to which of the two atti- tudes, A0 or A3, is likely to be more important will, of 25 course, vary with the particular attitudes and the particu- lar gs employed in any particular study. 4. The prediction of behavior based on knowledge of the attitude toward situation will be more accurate than predictions based on attitude toward object, but not as accurate as pre- dictions based on knowledge of the inter- action between attitude toward object and attitude toward situation, that is, r > r -B,A A -B,A —B,A o s s o The mean evaluation of the situation will be greater than the mean evaluation of the objects encountered in that situation, that is, MA >MA S O The mean perceived importance of the situation will be greater than the mean perceived im- portance of the objects encountered in that situation, that is, > M M(l-W) (W) METHOD Subjects and Procedure Data to test these hypotheses were gathered by means of a five page questionnaire. A total of 108 stu- dents, both males and females, enrolled in an intermediate psychology course (learning and motivation) at Michigan State University served as subjects (gs). In order to obtain an adequate sample size, the questionnaire was administered to two different groups of students. Both groups were enrolled in the same course, but during different terms. On both occasions the questionnaire was adminis- tered by the same experimenter (E), who was familiar to the gs, and identical procedures were followed. Respond- ing to the questionnaire was voluntary. No extra course credit was given for participating. All instructions written on the questionnaire were read, a section at a time, to the §5 as a group. Each section of the question- naire was completed by the entire group before going on to the next section. It was stressed by E that gs would re— main anonymous and that responses should be made with complete honesty. Each of the two sessions lasted approximately fifteen minutes. This included the reading of instructions 26 27 and the distribution and collection of the questionnaires. Explanations of the study and results obtained were later mailed to those gs who had indicated an interest. The Questionnaire The instructions of the first section of the questionnaire, and the format used were as follows. This research is designed to determine some of the factors which may affect the frequency of attending class in psychology courses. It is, of course, not possible to tell you the precise hypotheses being tested, for to do so would defeat our research pur- pose. I have, therefore, arranged for you to receive a handout later in the term which will tell you in more detail what the study is all about, and the re- sults obtained. In answering this questionnaire, there are no right or wrong answers, the best answer is your personal 0 inion. It is important that you make your judgments With complete honesty. Note that you will remain anonymous; you are asked not to write your name on this questionnaire. First, we would like you to list below the following information: (1) The courses which you have taken in psychology. List no more than five psychology courses. If you have taken more than five psychology courses, list the last five which you have taken. Do not list any course you are now taking, and do not list any course in which you had more than one instructor. (2) The number of class hours per week for each course. (3) For each course that you list, estimate the total number of times during the total quarter that, to the best of your recollection, you cut class. We are not interested in class cuts caused by illness, accident, unusual weather conditions, etc. Remember that you will remain anonymous. And remember also that our sole purpose is to find out what factors may affect class attendance. 28 The Professor in Number of Total Num- Title Each of These Class ber of of Courses Will Hence- Hours Per Cuts Per Course: forth Be Called: Week: Quarter: Course 1. Professor 1 Course 2. Professor 2 Course 3. Professor 3 Course 4. Professor 4 Course 5. Professor 5 It should be noted that behavior was not actually observed. The gs were asked to estimate the number of times they cut a class. There was no reason, however, to believe that the §S would not be able to give a reliable estimate or that they would not be honest in doing so. The gs were asked to list only psychology courses in an effort to keep the situation of attending classes as con- stant as possible. The second section of the questionnaire consisted of the following instructions and scale. It was this scale which provided measures of AS. We would next like to ascertain your general attitude toward attending class. We are ngt_interested in your attitude toward a particular instructor or toward a particular course, but only in how you feel about going to class in general. Please indicate how you 29 feel in general about the importance or unimportance of attending class by circling one number on the following scale. I feel that attending class is in general: C I RCLE ONE NUMBER 1 2 3 4 5 6 7 8 9 somewhat important very important to me important to me to me In a preliminary study, the first scale position was labeled "not at all important to me," and the fifth position was labeled "neither important nor unimportant." The ninth position was labeled as above. A very restricted range of responses resulted, with almost all §$ circling position 8 or 9. Given the nature of the sample, this is, in hindsight, not an unusual result. In order to increase the range of responses, the scale positions were therefore relabeled as above, with the result that the range, al- though still small, was increased to a degree large enough to allow meaningful calculations to be made. Instructions for the third section of the question— naire were as follows. You are now asked to indicate how you feel about £222 particular instructor that you had in each of the courses you‘listed on PAGE 1. We are interested only in how you feel about each particular instructor, PROFESSORS l, 2, 3, 4, and 5. Once again, remember that you will remain anonymous. And remember also that we are not in the least bit interested in finding out about the popularity of different instructors. We have no intention of identifying either you, or the instructors you rate, by name. 30 Five scales, one for each course, followed these instructions. The first of these scales was as follows. 1. My attitude toward the instructor that I have designated as PROFESSOR l (in COURSE l): CIRCLE ONE NUMBER 1 2 3 4 5 6 7 8 9 I disliked neither I liked instructor liked nor instructor very much disliked very much The other four scales were identical, except for the substi- tution of the appropriate numbers for the instructors. These five scales provided measures of A0. The final section of the questionnaire had the following instructions. Finally, we would like to know to what extent your going or not going to class was 'ointl determined by your attitude toward each of your instructors and_your general attitude toward attending class. For each course you listed on PAGE 1, you are asked to fill in a percentage following each of two simple scales. In each case, the two scales should add up to 100%. We are not interested in class cuts caused by illness, accident, unusual weather conditions, etc. The format for each of the five pairs of scales which followed was identical, except for appropriate numbering, to the following, which was the first one. 1. You had cuts (fill in from PAGE 1) in PRO- FESSOR 1's course. To what extent was the fre- quency of attendance or the frequency of cuts in this course determined by WRITE PER- a. your feelings about PROFESSOR 1? CENTAGE HERE % 0% 50% 100% 31 b. your general feeling about the importance or unimportance of attending class? 0% 50% 4100% c. Total percentage (a + b) 100% Analysis 2f the Data The 108 questionnaires were checked before any analysis of the data was carried out. Eight were excluded from the analysis because directions were not followed. Of these eight, five were excluded because the percentages in the last section of the questionnaire did not add up to 100%, and three were excluded because the SS had circled more than one number on a scale, or had omitted to circle a number. Nineteen questionnaires were excluded because S had listed less than three courses. It was felt that three courses would be the minimum number needed to pro- vide sufficient data. Each of the eighty-one §$ provided the following measures: (a) three to five measures of B, estimated number of class cuts, (b) an As' attitude toward attending class in general, (c) three to five measures of A0, atti- tudes toward instructors, and (d) three to five measures of AOAS, attitudes toward particular instructors within the context of attitude toward attending class in general. This last measure was derived by using the adaptation of Rokeach and Rothman's (1965) belief con- gruence model of cognitive interaction. Each A0 and AS 32 was weighted by the importance ratings indicated by the SS in the final section of the questionnaire (see formulas 4 and 5). For example, if a S had circled a 6 on the AS scale, a 3 on the scale for the second instructor (A02), and indicated that AS was 80% important in determining how many times he cut course 2, while A02 was 20% important, the derived value of Ao AS was 5.4. That is, 2 A02As (.80)6 + (.20)3 5.4. The derived AOAS measures are thus weighted averages and are numbers from 1 to 9. Rokeach and Rothman (1965) posit an additional comparison process, over and above the comparison of the relative importance of C and S with respect to one another in the context CS, when C reaches 100% in importance (and S 0%). This additional process is a further comparison of the relative importance of CS and C. There is the possi- bility that by virtue of the interaction between C and S within the framework of a person's total belief system his evaluation of CS may be even more extreme than his evalu— ation of C. That is, they deal with a comparison process sometimes referred to as overassimilation, a process whereby a stimulus not only takes on the valence of another stimu- lus with which it is associated, but in addition takes on an even stronger valence. Although some SS in the present study did rate either A0 or A8 as being 100% in importance, they were not asked to make further comparisons. It was felt that by doing so the questionnaire would become unduly lengthy 33 and complicated.8 Further, if overassimilation does occur it will work against the hypothesis that AoAs is a better predictor of B than is either A0 or A8. The data were analyzed in terms of three different methods of subdividing the sample. The first method in- cluded the total sample of eighty-one Ss. In this method, each S could have either three, four, or five measures of B, A0, and AoAs' The A0 and AOAS measures were averaged for each S, and mean value of each measure was used in the analysis. There was, of course, only a single AS for each S. Since the total sample was used in this method, the B for each S is referred to as B The B for each S is his T' T total number of estimated cuts divided by the total number of class hours per week. This preportion, rather than the absolute number of cuts, was used so that the 33 of differ- ent Ss would be comparable. The second method resulted in a subsample of seventy-one Ss. In this subsample, only one measure of B, A0, AS, and AOAS per S was used. The B for each S is referred to as B1, and is simply S's estimated number of cuts, since class hours per week for each S was held con- stant at three. If S had listed more than one course which had three class hours per week, the first one listed 8See Insko (1967, pp. 159-160) for a discussion of the conceptual and procedural problems involved in Rokeach and Rothman's (1965) approach to the treatment of over- assimilation effects. 34 was arbitrarily chosen to be included in the analysis. Ten Ss had not listed a course with three hours per week. In this subsample, therefore, it was the number of class hours per week that was held constant for each S. Courses, however, varied across Ss. The third method resulted in five subsamples of varying sizes. This method was like the second in that only one measure of B, A0, AS, and AOAS per S was used. The B's in these five subsamples are referred to as 32' B B3, B and B6' and are simply the estimated number of 4' 5’ cuts. In each of these five subsamples, there is one specific course per S (that is, either the course numbered 200, or the course numbered 225, and so on). The course, and thus the class hours per week, are held constant for each of the Ss within a subsample. But courses do, and class hours per week can, vary across subsamples. RESULTS Subsample ST Subsample B T includes the entire sample of eighty- one Ss. ET for each S'is his total number of estimated cuts divided by his total number of class hours per week. There are from three to five courses for each S, and class hours per week vary. The results for subsample BT are shown in Tables 1 and 2. Table 1 shows the ranges, means, variances, and standard deviations of the obtained measures for this sub- sample. The range of B is from 0.00 to 5.17. This range T corresponds to those Ss who indicated zero number of cuts (N = 8) at the lower extreme, to one S who indicated his total number of estimated cuts to be sixty-two and his total number of class hours per week to be twelve, at the upper extreme. That the obtained distribution is skewed is indicated by the mean BT of 1.02. The variance of BT is 1.15. There is little variability in class-cutting behavior. The range of A0 is from 1.67 to 9.00, and the range of both AS and AOAS is from 1.00 to 9.00. These distributions are also skewed, with the most skewness occurring for AS and the least for A0. The mean AOAS 35 36 TABLE l.--Ranges, means, variances, and standard deviations of obtained measures for subsample B T. Measure Statistic a b b b b BT Ao AS AOAs (w) (l-w) Range 0.00- 1.67- 1.00- 1.00- 0.00— 0.00- 5.17 9.00 9.00 9.00 1.00 1.00 Mean 1.02 5.86 7.05 6.66 .37 .63 Variance 1.15 2.10 3.88 2.95 .05 .05 Standard Deviation 1.07 1.45 1.97 1.72 .23 .23 Note: N = 81. aBT for each S is his total number of estimated cuts divided by his total number of class hours per week. There are from three to five courses for each S, and class hours per week vary. bMeasures for each S are means. 37 (6.66) is intermediate between that of A0 (5.86) and of AS (7.05). In other words, the mean evaluation of the situation (AS) in this subsample is greater than the mean evaluation of the objects (A0) in that situation (39 = 4.4, p < .001), thus supporting hypothesis 4. Since the measures of A0 and AOAs for each S are means, the variance of the As distribution (3.88) is greater than that of either the A0 (2.10) or the AoAs (2.95) distribution. Summary statistics of the obtained importance ratings for subsample BT are shown in the last two columns of Table l. The headings (w) and (l-w) refer to the im- portance of feelings about the instructor and the importance of feelings about attending class in general, respectively. The range of both (w) and (l-w) is from 0.00 to 1.00. Differences between mean (w) and mean (l-w), .37 and .63 respectively, may be compared with the differences between mean A0 and mean AS. Both differences indicate that, in this instance, the situation is considered by the Ss to be more important than the object. The mean per— ceived importance of the situation is greater than the mean perceived importance of the objects encountered in that 9No tests for homogeneity of variance were carried out and thus this E value, as well as all other Es reported in this study were not corrected for any heterogeneity of variance. Hays (1963, p. 322) points out that for rela- tively large samples of equal size conclusions derived from a t test will be little affected when variances are not homogeneous. Boneau (1960) also states that the value of E is not markedly affected, except by rather strong vio- lations, unless sample size is very small. 38 situation in the determination of behavior (3 = 5.1, p < .001). This result thus supports hypothesis 5. Since a S's indication of (w) is always perfectly correlated (-1.00) with his indication of (l-w), the variances (.05) and standard deviations (.23) of these two distributions are identical. The relevant intercorrelations between the obtained measures for subsample BT are shown in Table 2. All £5 reported in this study are Pearson product-moment corre- lations. The ES between the behavioral and attitudinal measures are negative because of the way in which the attitude scales were numbered. This holds true for all subsamples in the study. The obtained £5 between B and A0, AS, and AOAS T are -.20, -.46, and -.61, respectively. The E of -.20 between BT and A0 is not significant (3 = 1.6). Both AS and AOAS are, however, significantly correlated with BT' The E of -.46 between B and As is significant (5 = 4.6, T p < .001), but less than the E of -.61 between BT and the AOAS combination (5 = 6.8, p < .001). r and r is -—B,AS -B,AO statistically significant (3 = 2.00, p < .05, two-tailed), The difference between as is the difference between and r (E’= 3.84, r -B’AOAS “BIAO p < .001, two—tailed), and the difference between rB A A and EB,A (E = 3.00, p < .01, two-tailed). These results 3 thus support hypothesis 4 which stated that the prediction of behavior based on knowledge of the attitude toward 39 TABLE 2.--Intercorrelations between the obtained measures for subsample BT‘ Measure Measure BTa Ab A AAb O S OS a BT 1.00 A b -.20 1.00 0 As -.46*** .07 1.00 AOASb -.6l*** .36*** .84*** 1.00 Note: N = 81. aBT for each S is his total number of estimated cuts divided by his total number of class hours per week. There are from three to five courses for each S, and class hours per week vary. b Measures for each S are means. ***p < .001 40 situation will be more accurate than predictions based on attitude toward object, but not as accurate as predictions based on knowledge of the interaction between attitude toward object and attitude toward situation. The multiple correlation (3), using B as the criterion variable and A0 and AS as the predictor vari- ables, is .49. This is lower than the E of .61 between B and AOAS, thus supporting hypothesis 3. The £3 between A0 and AOAS (.36) and between AS and AOAS (.84) are, of course, high and also significant EAo'AoAs and EAs'AoAs is statistically significant (t = 6.98, p < .001, two- (p < .001). The difference between tailed). That A0 and A8 are two different types of atti— tudinal measures is indicated by the obtained £_between these measures of .07, which is close to zero. This sup- ports hypothesis 2, which stated that AC and AS would be nonsignificantly correlated. Obtained coefficients of determination (£25) indi- cate that the proportion of the variance in BT accounted for by A0, A8, and AOAS is .04, .21, and .37, respectively. This is another indication that AOAS is a better predictor of B than is either A0 or A5. The proportion of the vari— ance in AOAS accounted for by A0 is .13, and by As is .71. This reflects the greater importance attributed by the Ss to AS than to A0. Correlations involving (w) and (l—w) were calcu- lated, but are not considered to be psychologically 41 relevant. The only £5 involving these two measures which may be of interest are those between A0 and (w) and be- tween AS and (l-w). In this subsample, these 55' -.02 and .02, respectively, are close to zero and are nonsignificant, as are all other £5 between these two measures and the be- havioral and attitudinal measures. Subsample S1 The results for subsample B (N = 71) are shown in 1 Tables 3 and 4. B1 for each S’is his estimated number of cuts. Class hours per week are constant at three. There is one course per S, and different courses across Ss. Table 3 shows the ranges, means, variances, and standard deviations of the obtained measures for this sub- sample. The range of B1 is from zero to 20. The mean of 2.83 indicates that this distribution is skewed, as was the distribution of B . Due to the differences in range T for B1 and BT’ the variances of these two distributions differ markedly. The variance of the B1 distribution is 13.83, while the variance of the BT distribution is 1.15. The range of each of the three attitudinal measures is from 1.00 to 9.00. These distributions are also skewed, with the most skewness occurring for AS and the least for A . As in subsample B the mean evaluation of the situ— o T' ation (7.06) is greater than the mean evaluation of the objects (6.11) in that situation (2 = 2.7, p < .01). Note the similarity of the mean As for subsample B (7.06) with l 42 TABLE 3.-—Ranges, means, variances, and standard deviations of obtained measures for subsample B1. Measure Statistic a B1 A0 AS AOAs (w) (l-w) Range 0-20 1.00- 1.00- 1.00- 0.00- 0.00- 9.00 9.00 9.00 1.00 1.00 Mean 2.83 6.11 7.06 6.81 .38 .62 Variance 13.83 6.38 3.74 3.88 .10 .10 Standard Deviation 3.72 2.53 1.93 1.97 .31 .31 Note: N = 71. aB for each S is his estimated number of cuts. Class hourE per week Is held constant at three. There is one course per S, and courses vary across Ss. 43 the mean AS for subsample B (7.05). In both instances, T the situation is highly evaluated, and also evaluated higher than the objects in the situation. Since there is only one A0 measure for each S in this subsample, the variance of the AO distribution (6.38) is larger than that of the variance of the AO distribution for subsample B (2.10), where means were used. For sub- T sample Bl the variance of AS is 3.74 and the variance of A A is 3.88. o 5 Summary statistics of the obtained importance ratings'for subsample B1 are shown in the last two columns of Table 3. Both (w) and (l-w) range from 0.00 to 1.00. The means of .38 and .62 are almost identical to those of .37 and .63 obtained for subsample B Again, the mean T' perceived importance of the situation is greater than the mean perceived importance of the objects encountered in that situation (5 = 3.4, p < .001). The variance of both (w) and (l-w) is .10. Relevant intercorrelations between the obtained measures for subsample B1 are shown in Table 4. All three correlations between the behavioral measure and the three attitudinal measures are significant. The 5 of -.56 (E_5 5.6, p < .001) between B1 and AOAs is higher than either the £,of -.38 (E = 3.4, p < .01) between B1 and A0 or the £,of -.35 (E = 3.1, p < .01) between B and AS. 1 The difference between EB’AoAs and £B,AO 1n this subsample is statistically significant (E = 1.76, p < .10, 44 TABLE 4.--Intercorrelations between the obtained measures for subsample Bl. Measure Measure Bla A A A A o s o s a B1 1.00 A -.38** 1.00 0 AS -.35** .00 1.00 A A -.56*** .50*** .66*** 1.00 o 5 Note: N = aB for each S is his estimated number of cuts. Class hours per week Is held constant at three. There is one course per S, and courses vary across Ss. **p < .01 ***p < .001 45 two-tailed), as is the difference between rB A A and — ’ o 3 EB A (E = 2.53, p < .02, two—tailed). The best pre- I s dictor of B is thus AOAS even though both A0 and AS are significantly correlated with B. There is no statistically significant difference between r and r (t = 0.20, —B,AO —13,As — NS). The §_value in this subsample is .52, and thus is lower than the 5 of .56 between B and AOAS. As in subsample B the 55 between A0 and AoAs and TI between A8 and AOAS are high and statistically significant (p < .001). The difference between and r £A,AA —A,AA o o s s o s is statistically significant (E = 1.67, p < .10, two- tailed). Also as in subsample B there is a zero corre- Tr lation between A0 and AS. Obtained £28 indicate that the proportion of the variance in B1 accounted for by A0, As' and AOAS is .14, .12, and .31, respectively. This again indicates that AOAS is a better predictor of B than is either A0 or A5. The proportion of the variance in AOAS accounted for by A0 is .25 and by As is .44, indicating the greater im- portance attributed to AS than to A0. The obtained as between A0 and (w) and between AS and (l-w), .13 and .02, respectively, are statistically nonsignificant, as are all other £5 between these two measures and the behavioral and attitudinal measures in this subsample. 46 Subsamples S2, S3, S4, S5, §§§.§5 Results for subsamples B2 through B6 are presented in Tables 5 through 14. The B in each of these subsamples is simply the estimated number of cuts. There is one course per S. The course, and thus the number of class hours per week, is the same for all Ss within a particular subsample. These subsamples are, of course, smaller than the two previously discussed subsamples, but the results obtained may be viewed as checks on the stability and re- and B . Be- T 1 cause of the small sizes of these subsamples, only the liability of those obtained in subsamples B summary statistics and the relevant correlations were calculated. Subsample S2.-—The results for subsample 32 (N = 35) are shown in Tables 5 and 6. Table 5 shows the ranges, means, variances, and standard deviations of the obtained measures for this subsample. The range of B2 is from zero to 14. The mean of 1.86 reflects the skewness of this distribution. The variance of B is 6.18, and the standard 2 deviation is 2.48. The range of A0 is from 1.00 to 9.00, while that of As is from 2.00 to 9.00, and that of AoAs is from 3.50 to 9.00. The mean evaluation of the situation (7.28) is greater than the mean evaluation of the objects (6.34) in the situation. All three of these distributions are skewed. Note that the mean of AOAs (7.36) is larger than that of either A0 (6.34) or AS (7.28). This occurs 47 TABLE 5.--Ranges, means, variances, and standard deviations of obtained measures for subsample B 2. Measure Statistic a B2 A0 AS AOAS (w) (1-w) 0-14 1.00- 2.00- 3.50- 0.00- 0.00- Range 9.00 9.00 9.00 1.00 1.00 Mean 1.86 6.34 7.28 7.36 .39 .61 Variance 6.18 5.77 2.72 1.83 .11 .11 Standard Deviation 2.48 2.40 1.65 1.35 .33 .33 Note: N = 35. a32 for each S is his estimated number of cuts. There is one course per S. The course, and thus the number of class hours per week, is the same for all Ss. 48 because the derived AoAs measures are weighted averages, not simple means. The variance of A0 (5.77) is larger than that of either As (2.72) or of AOAS (1.83). Again, there is more variability in evaluations of course instructors than in evaluations of the general importance of attending class. Summary statistics of the obtained importance ratings for subsample B2 are shown in the last two columns of Table 5. Both (w) and (l-w) range from 0.00 to 1.00. The means of .39 and .61 are quite similar to those of the two previously discussed subsamples. The mean perceived importance of the situation is greater than that of the objects in that situation. The variances of (w) and (l-w) are .11. Relevant intercorrelations between the obtained measures of subsample B2 are shown in Table 6. The ob- tained E of -.31 (E = 1.9) between 82 and A0 is not signifi- cant. The 5 of -.34 (E = 2.1, p < .05) between B and AS 2 is significant, but is not as high as the E of -.55 (E = 3.8, p < .001) between B2 and AOAS. The best pre- dictor of B is AOAS. The next best predictor is AS, and the poorest predictor is A0. In this subsample too, 3 (.48) is lower than the 5 between B and AOAS (.55). The £3 between A0 and AOAS (.38) and between A8 and AOAS (.64) are both significant (p < .05, and p < .001), respectively). Again, the correlation between A0 and AS (-.07) is essentially zero. 49 TABLE 6.--Intercorrelations between the obtained measures for subsample B2. Measure Measure 13a A A AA 2 o s o s a B2 1.00 A -.31 1.00 0 AS -.34* -.07 1.00 A A -.55*** .38* .64*** 1.00 o s Note: N = 35. a82 for each S is his estimated number of cuts. There is one course per S. The course, and thus the class hours per week, is the same for all Ss. *p < .05 ***p < .001 50 Obtained £23 indicate that the proportion of the variance in B accounted for by A , A , and A A is .10, 2 o s o s .11, and .30, respectively, reflecting the fact that AOAS is the best predictor of B. The prOportion of the vari- ance in AOAS accounted for by A0 is .15, and by As is .42. Subsample S3.--Results for subsample B3 (N = 30) are shown in Tables 7 and 8. Table 7 shows the ranges, means, variances, and standard deviations of the obtained measures for this subsample. The range of B3 is from zero to 25. The skewness of the distribution of B3 is indicated by the mean of 4.13. The variance of B is 25.18, and the 3 standard deviation is 5.02. The range of A0 is from 1.00 to 9.00, while that of As is from 2.00 to 9.00, and that of AOAS is from 3.00 to 9.00. The mean of A0 (4.97) is the lowest of the three, and is essentially at the midpoint of the nine point scale. The mean of AS, the largest of the three, is 6.87, and the mean of AOAS is 6.05. Again, the mean evaluation of the situation is greater than that of the objects in the situation. The variance of A0 (6.30) is larger than that of either AS (4.12) or AOAS (4.29). The last two columns of Table 7 show the summary statistics of the obtained importance ratings for sub- sample B Both (w) and (l-w) range from 0.00 to 1.00. 3. The mean of (w) is .38 and the mean of (l—w) is .62, figures which are very close to those of three previously discussed subsamples. As in the other subsamples, the 51 TABLE 7.--Ranges, means, variances, and standard deviations of obtained measures for subsample B 3. Measure Statistic a B3 A0 AS A AS (w) (l w) 0-25 1.00- 2.00— 3.00~ 0.00~ 0.00- Range 9.00 9.00 9.00 1.00 1.00 Mean 4.13 4.97 6.87 6.05 .38 .62 Variance 25.18 6.30 4.12 4.29 .08 .08 Standard Deviation 5.02 2.51 2.03 2.07 .29 .29 Note: N = 30. aB3 for each S is his estimated number of cuts. There is one course per 8. of class hours per week,_is the same for all Ss. The course, and thus the number 52 mean perceived importance of the AS is greater than the mean perceived importance of A0. The variances of (w) and (1-w) are .08. Relevant intercorrelations between the obtained measures of subsample B3 are shown in Table 8. Neither the obtained 5 of -.16 (E = 0.9) between B3 and A0 nor the E of -.29 (E = 1.6) between B3 and AS is significant, although As is a better predictor than is A0. The E of -.39 (E = 2.3, p < .05) between B3 and AOAs is significant. It is not as high as those in the three previously dis— cussed subsamples, but is nevertheless the best predictor of B. It is also higher than the obtained 5 of .33. The £5 between A0 and AOAs (.56) and between AS and AOAS (.57) are both significant at the .01 level. As in the previous subsamples, the correlation between A0 and As (.04) is essentially zero. Obtained £25 indicate that the proportion of the variance in B3 accounted for by A0, As, and AOAS is .02, .09, and .15, respectively, reflecting the fact that AOAS is the best predictor of B. The proportion of the vari- ance in AOAS accounted for by A0 is .31, and by As is .33. Subsample S4.--Tables 9 and 10 show the results for subsample B4 (N = 29). The ranges, means, variances, and standard deviations of the obtained measures for this subsample are shown in Table 9. The range of B4 is from zero to 15. The mean of 2.72 again indicates the skewness 53 TABLE 8.--Intercorre1ations between the obtained measures for subsample B3' Measure Measure 33a A A A A O S O S a B3 1.00 A -.16 1.00 0 As -.29 .04 1.00 A A -.39* .56** .57** 1.00 O S Note: N = 30. aB3 for each S is his estimated number of cuts. There is one course per S. The course, and thus the class hours per week, is the same for all Ss. *p < .05 j **p < .01 54 TABLE 9.-—Ranges, means, variances, and standard deviations of obtained measures for subsample B 4. Measure Statistic a B4 A0 As AoAs (w) (l-w) 0-15 1.00- 2.00— 2.00- 0.00- 0.00- Range 9.00 9.00 9.00 1.00 1.00 Mean 2.72 6.28 7.52 7.30 .35 .65 Variance 17.44 6.13 3.49 2.93 .08 .08 Standard Deviation 4.18 2.48 1.87 1.71 .28 .28 Note: N = 29. aB4 for each S is his estimated number of cuts. There is one course per S. The course, and thus the number of class hours per week, 15 the same for all Ss. 55 of the behavioral measure. The variance of B is 17.44, 4 and the standard deviation is 4.18. The range of A0 is from 1.00 to 9.00, while the range of both AS and AoAs is from 2.00 to 9.00. The mean of A0 (6.28) is lower than that of either As (7.52) or of AOAS (7.30). Mean evaluations of the situation are again higher than mean evaluations of the objects. The variance of A0 (6.13), however, is greater than that of either As (3.49) or AOAS (2.93). Summary statistics of the obtained importance ratings for subsample B4 are shown in the last two columns of Table 9. Both (w) and (l-w) range from 0.00 to 1.00. The mean of (w) is .35, and .65 is the mean of (l-w). That is, the mean perceived importance of the situation is greater than that of the objects. The variances of (w) and (l-w) are .08. Table 10 shows the relevant intercorrelations be- tween the obtained measures of subsample B4. The obtained 5 of -.24 (E = 1.3) between B4 and A0 is not significant. The E of -.45 (E = 2.7, p < .05) between B4 and AS is significant, but not as large as the 5 of -.57 (E = 3.7, p < .01) between B4 and AOAS. Once again, AOAS is the best predictor of B. The next best predictor is AS, and the poorest predictor is A0. The obtained 3 (.51) is lower than the £.°f .57 between B and AOAS. The correlations between AOAs and its two com- ponents are both significant, but the E of .36 (p < .05) TABLE 10.--Intercorrelations for subsample B4. 56 between the obtained measures Measure Measure 134":1 A A A A o s s a B4 1.00 A -.24 1.00 0 AS -.45* -.01 1.00 A A -.57** .36* .81*** 1.00 o 5 Note: N = 29. aB4 for each S is his estimated number of cuts. There is one course par S. The course, and thus the class hours per week, is the same for all Ss. *p < .05 **p < .01 ***p < .001 57 between A0 and AOAs is not as large as the E of .81 (p < .001) between AS and AOAS. The correlation between A0 and AS (-.01) is essentially zero. Another indication that AOAS is the best predictor of B is shown by the fact that the proportion of the vari- ance in B4 accounted for by A0, A8, and AOAs is .06, .20, and .32, respectively. The prOportion of the variance in AOAS accounted for by A0 is .13, and by As is .65, reflect- ing the greater evaluation of AS than of A0. (N = 25) Subsample S5.--The results for subsample B5 are shown in Tables 11 and 12. Table 11 shows the ranges, means, variances, and standard deviations of the obtained measures for this subsample. The range of B5 is from zero to 25. This distribution of behavioral measures is skewed, as indicated by the mean of 4.56. The variance of B5 (38.01) is the largest of all the subsamples, and is due in part to the small N. The range of both A0 and As is from 2.00 to 9.00, while AOAs has a range of from 3.20 to 9.00. Of the three attitudinal measures, AS has the highest mean (6.64), and A0 has the lowest mean (5.60). The mean evaluation of the situation is again greater than the mean evaluation of the objects in that situation. The mean of AOAs is 6.49. The variance of A0 is 5.36, while the variance of AS is 3.99, and that of AoAs is 3.77. The last two columns of Table 11 show summary statistics for the obtained importance ratings in subsample 58 TABLE ll.--Ranges, means, variances, and standard deviations of obtained measures for subsample B5. Measure Statistic a B5 A0 A8 AOAS (w) (l-w) 0—25 2.00- 2.00- 3.20- 0.00- 0.25- Range 9.00 9.00 9.00 0.75 1.00 Mean 4.56 5.60 6.64 6.49 .30 .70 Variance 38.01 5.36 3.99 3.77 .04 .04 Standard Deviation 6.16 2.32 1.99 1.94 .21 .21 Note: N = 25. aBS for each S is his estimated number of cuts. There is one course per S. The course, and thus the number of class hours per week Is the same for all Ss. 59 BS’ The range of (w) is from 0.00 to 0.75, and the range of (l-w) is from 0.25 to 1.00. The means of (w) and (1-w), .30 and .70, respectively, are the most extreme of all the subsamples, and show the greater perceived importance of the situation. The variance of (w) and (l-w) is .04. Relevant intercorrelations between the obtained measures of subsample B5 are shown in Table 12. None of the correlations between B5 and the attitudinal measures is significant. The 5 of -.38 between B and AOAS (E = 2.0), 5 however, is larger than that of -.25 between B and A0 5 (E = 1.3) or that of -.27 between B and AS (E.= 1.4). 5 Thus, even though AOAS is not significantly correlated with B, it is nevertheless a better predictor than either A0 or A8. The 5 between B and AoAs (.38) is also higher than the obtained 3 of .33. Both the S’of .58 (p < .01) between A0 and AOAS and the a of .84 (p < .001) between A5 and AOAS are signifi- cant. The £_of .20 between A0 and As is large as compared with those of the other subsamples, but is not significant. The proportion of the variance in B5 accounted for by A0, AS, and AOAS is .06, .07, and .14, respectively. The prOportion of the variance in AOAs accounted for by A is .34, and by As is .71. O Subsample S6.--Tables 13 and 14 show the results for subsample B6 (N = 20). Ranges, means, variances, and standard deviations of the obtained measures for subsample B are shown in Table 13. The range of B is from zero 6 6 60 TABLE 12.--Intercorrelations between the obtained measures for subsample BS’ Measure Measure a B5 A0 A3 AoAs a 85 1.00 A -.25 1.00 0 As -.27 .20 1.00 A A -.38 .58** .84*** 1.00 o s Note: N = 25. aBS for each S is his estimated number of cuts. There is one course per S. The course, and thus the class hours per week, is the same for all Ss. **p < .01 ***p < .001 61 TABLE l3.—-Ranges, means, variances, and standard deviations of obtained measures for subsample B 6. Measure Statistic a B6 A0 A8 AOAs (w) (1-w) 0-12 2.00— 3.00- 3.40— 0.00- 0.10- Range 9.00 9.00 9.00 0.90 1.00 Mean 3.40 6.25 7.15 6.96 .43 .57 Variance 8.94 6.89 2.73 2.62 .11 .11 ~ Standard Deviation 2.99 2.62 1.65 1.62 .33 .33 Note: N = 20. aB6 for each S is his estimated number of cuts. There is one course par S. The course, and thus the number of class hours per week, is the same for all Ss. 62 to 12, and the skewness of the distribution is indicated by the mean of 3.40. The variance of B is 8.94. 6 The range of A0 is from 2.00 to 9.00. AS has a range of from 3.00 to 9.00, and AoAs has a range of from 3.40 to 9.00. The means of A0, A5, and AOAS are 6.25, 7.15, and 6.96, respectively, indicating that the mean evaluation of the situation is once again greater than the mean evaluation of the objects. AC has the largest vari- ance (6.89), while the variance of As is 2.73 and that of AOAS is 2.62. Summary statistics of the obtained importance ratings for subsample B6 are shown in the last two columns of Table 13. The mean of (w) is .43, and (l-w) has a mean of .57. As in the other subsamples, the mean perceived importance of the situation is greater than that for the objects irxthe situation. The variance of (w) and (l-w) is .11. Table 14 shows the relevant intercorrelation be- tween the obtained measures for subsample B6' The E of -.68 between B and A0 is significant (E = 4.0, p < .001). 6 The E of .13 between B6 and AS, however, is not significant (3 = 0.5). This is the only subsample in which A0 is a better predictor of B than is As. In part, this may be attributed to the small sample size. It may also, how- ever, be due to the known extreme popularity of the pro— fessor who generally teaches this particular course. This result indicates that AO may be a better predictor of B in 63 TABLE l4.--Intercorrelations between the obtained measures for subsample B6' Measure Measure B a A A A A 6 o s o s a A -.68*** 1.00 0 AS .13 -.26 1.00 A A -.78*** .79*** .08 1.00 o s Note: N = 20. aB for each 8 is his estimated number of cuts. There is one course per S. The course, and thus the class hours per week, is the same for all Ss. ***p < .001 64 spite of the fact that the mean evaluation of the situation and the mean perceived importance of the situation are both greater than for the objects in the situation. Note that the correlation between B6 and AS (.13) is positive. That is, it is in the opposite direction of what would be expected. The £_between B6 and AOAS is -.78, and is significant (E_= 5.5, p < .001). It is also higher than the E of -.68 between B and A0. Thus, although A0 6 is a good predictor of B in this subsample, AOAS is better. The obtained 3 in this subsample is .68, and thus lower than the £_between B and AOAS (.78). The correlation of .79 between A0 and AOAS is significant (p < .001), while the E of .08 between AS and AOAS is close to zero. A second reversal of expected direction in the correlations for this subsample is the E of -.26 between A0 and AS. Although relatively high, this 5 is not significant. Obtained £23 indicate that the proportion of the variance in B6 accounted for by A0, A8, and AOAS is .46, .02, and .61, respectively, reflecting the fact that AOAs is the best predictor of B. The proportion of the vari- ance in AOAS accounted for by A0 is .63, while that accounted for by AS is only .01. 65 Summary 9: Results 12 Relation Eg_t e Hypotheses Summary 2: results pertaining £2.50! AS, and A055 2E predictors 9E S.--The first four hypotheses were con- cerned with the accuracy with which E could be predicted. The main results concerning these four hypotheses are summarized in Table 15 for each of the subsamples. Al— though subsamples B1 through B6 may be viewed as repli- cations, in a sense, of subsample B it should be kept in TI mind that an unknown degree of dependency may exist among the correlational values presented in Table 15. The ob— tained values for a member of subsample B for example, 5’ might also be included in subsample B and so on. Caution 2, must therefore be exercised when interpreting the results, taken as a whole, that are presented in Table 15. Hypothesis 1 stated that the prediction of behavior will be more accurate when the interaction between atti- tude toward object and attitude toward situation is taken into account than when only attitude toward object or to- £B,AOAS > £B,AO or Columns four, five and six of Table 15 show that: ward situation is considered, that is, rB,A . s (a) The ES between B and AOAS are significant in six out of seven subsamples. The Es between B and A0, however, are significant in only two of seven cases, and the ES between B and A8 are significant in three of seven cases. (b) In each of those subsamples where the correlation be- tween B and A0 (B6) or between B and AS (BT81), or both (Bl) 66 TABLE 15. --Summary of results pertaining to A0 ,A3, and A OSA as predictors of B. Coefficient of Correlation Subsample N r R AO,AS B,AO B,AS B,AOAS B-AO,AS BTa 81 .07 -.20 -.46*** -.6l*** .49 Blb 71 .00 -.38** -.35** -.56*** .52 82C 35 -.07 -.31 -.34* -.55*** .48 B3C 30 .04 —.16 -.29 -.39* .33 B4C 29 -.01 -.24 -.45 -.57** .51 BSC 25 .20 -.25 -.27 -.38 .33 86° 20 -.26 -.68*** .13 —.78*** .68 Note: N refers to sample size. aBT for each S is his total number of estimated cuts divided by his total number of class hours per week. There are from three to five courses for each S, and class hours per week vary. bBl for each S is his estimated number of cuts. Class hours per week are constant at three. There is one course per S, and different courses across Ss. CB for each S is his estimated number of cuts. There is one course per S. The course, and thus the number of class hours per week, is the same for all Ss within a particular subsample. — *p < .05 **p < .01 ***p < .001 67 are significant, the correlation between B and AOAS is larger. (c) Even in the subsample (B5) where none of the correlations between B and the attitudinal measures is significant, the 5 between B and A0 or between B and AS. Thus hypothesis 1 is supported in every one of the seven subsamples in the study. Hypothesis 2 stated that measures of attitude to- ward object and of attitude toward the situation in which the object is encountered will be independent of one another, that is, EA A z 0. The third column of Table 15 o’ 3 shows that A0 and AS are nonsignificantly correlated in each of the seven subsamples, and that in most instances the E is essentially zero. ‘Hypothesis 3 stated that the prediction of behavior will be as accurate, or more accurate, when the adaptation of the belief congruence configurationist model is employed than when the additive model of multiple correlation is £B,A A > BB-A ,A ’ o s o ~s this hypothesis are shown in the last two columns of Table used, that is, Results pertaining to 15. Hypothesis 3 is supported in all of the seven sub- samples. Hypothesis 4 stated that the prediction of behavior based on knowledge of the attitude toward situation will be more accurate than predictions based on attitude toward object, but not as accurate as predictions based on knowl- edge of the interaction between attitude toward object and . . . . > attitude toward Situat1on, that 18, £B,AOA > £B,A EB,A 68 Columns four, five, and six of Table 15 show that the rank order of the correlations between B and the three atti- tudinal measures, from highest to lowest, is AOAS, AS, A0 in five of the seven subsamples (BT, B2, B3, B4, and B5), and AOAS, A0, AS in two of the five subsamples (B1 and B6). Thus hypothesis 4 is supported in five of the seven sub- samples. Summary 9E results pertaining Eg_mean evaluation and mean perceived importance 9: situation and objects.-- Hypotheses 5 and 6 were concerned with the mean evaluation of the situation and the objects in that situation, and with the mean perceived importance of the situation and objects. The main results concerning these two hypotheses are summarized in Table 16 for each of the subsamples. Just as for the results presented in Table 15, the possi- bility of dependencies among the values in Table 16 must be kept in mind when interpreting the results taken as a whole. Hypothesis 5 stated that the mean evaluation of the situation will be greater than the mean evaluation of the objects encountered in that situation, that 13, MA > MA . s o The third and fourth columns of Table 16 show that this hypothesis is supported in every one of the seven sub- samples. For subsample B the difference between mean TI AS and mean A0 is significant (E = 4.4, p < .001). There is also a significant difference between mean AS and mean A0 in subsample B1 (E = 2.7, p < .01). Significance tests 69 TABLE l6.—-Summary of results pertaining to mean evaluation and mean perceived importance of situation and objects. Mean Subsample N A5 A0 (l-w) (w) BTa 81 7.05 5.86 .63 .37 b 81 71 7.06 6.11 .62 .38 82° 35 7.28 5.77 .61 .39 83° 30 6.87 4.97 .62 .38 B4C 29 7.52 6.28 .65 .35 85° 25 6.64 5.60 .70 .30 86° 20 7.15 6.25 .57 .43 Note: N refers to sample size. aThe difference between mean As and mean A0 is significant at the .001 level (t = 4.4, two-tailed, with 80 degrees of freedom). The difference between mean (l-w) and mean (w) is also significant at the .001 level (E = 5.1, two-tailed, with 80 degrees of freedom). bThe difference between mean As and mean A is significant at the .01 level (3 = 2.7, two-tailed, with 70 degrees of freedom). The difference between mean (l-w) and mean (w) is also significant at the .01 level (E = 3.4, two-tailed, with 70 degrees of freedom). CE tests were not carried out on these subsamples because of the small N. 70 were not carried out on the other means in columns three and four because of the small N. Hypothesis 6 stated that the mean perceived im- portance of the situation will be greater than the mean perceived importance of the objects encountered in that situation, that is, M(l-w) > M(w)° This hypothesis is supported by results for all seven subsamples, as shown in the last two columns of Table 16. The difference between mean (l-w) and mean (w) is significant for both subsample B (E = 5.1, p < .001) and for subsample B1 (E.= 3.4, T p < .01). DISCUSSION The basic or underlying question in studies of the relationships between attitudes and behavior is whether or not knowledge of an individual's expressed attitudes will aid in the prediction of his overt behavior. There are at least three conditions which must be met when attempting to answer this question. Investigators must first of all derive their experimental hypotheses from some prior con- ception of the attitude construct. The clearer the con- ceptualization, the more meaningful are the results of the study likely to be. A second condition is to ade— quately assess the attitudes under investigation. The type of measurements made will depend both upon the par— ticular conceptualization of attitude which is employed and upon the use to which this information is to be put when making predictions about behavior. A third condition is to adequately assess the interrelationships between attitudes and behavior. The form of this latter measure will depend to a large extent upon the manner in which the first two conditions are met, but will also depend upon other factors such as the type of behavior under study, time relationships involved, and so on. 71 72 The purpose of this section is to discuss these conditions in relation to Rokeach's (1968) definition of attitude and in relation to the hypotheses, design, and results of the present study. What are some of the impli- cations of the conceptual definition from which the hy- potheses were derived? What are some of the implications of the results of the present study for both previous research and for future research on the relationships be- tween attitudes and behavior? Limitations in methodology must also be kept in mind when attempting to generalize from the results of the present study. Some of these limitations, and suggestions for their resolution, will also be discussed. The Conceptualization giiAttitude Two major components of Rokeach's (1968) definition of attitude were tested in the present study. The first is that an attitude is a predisposition to respond in a preferential manner. This aspect of the definition is, of course, fundamental to the hypothesis that an indi- vidual's behavior can be predicted from knowledge of his attitudes. The other component of the definition which was tested is that attitudes may be toward situations as well as toward objects. From this second aspect of the definition is derived the hypothesis that whenever an attitude object is encountered in some situation, at least two attitudes, A0 and As, will be activated. Behavior can 73 thus be predicted on the basis of knowledge of the outcome of the cognitive interaction between A0 and As. Rokeach's conceptualization of attitude as a set of interrelated predispositions to respond in a preferen- tial manner differs from that of some other theorists. Horowitz (1944), for example, views an attitude not as a predisposition to respond but as a response. An attitude is conceptualized by Doob (1947) as an implicit response. Although not expressly excluding predispositional aspects of an attitude, Campbell (1950) and Green (1954) both employ the attitude construct to refer to consistency among a set of responses. As Rokeach points out, however, most writers . . . seem to agree that an attitude is a predisposi- tion of some sort, although there seems to be some difference of Opinion about what kind of predisposi- tion it is: . . . to respond; . . . to evaluate; . . . to experience; . . . to be motivated, . . . to act. In the present formulation we prefer simply 'predisposition to respond' with the understanding that a response may be either a verbal expression of an opinion, or some form of non-verbal behavior (1968, p. 453). By emphasizing that an attitude is a "predisposi- tion to respond," Rokeach's (1968) definition avoids the implication that the response itself is either affective or evaluative. It may be either, or it may be a combi- nation of both. The main point, however, is that the use of this conceptualization of attitude leads to the hy— pothesis that a person's behavior toward an object can be predicted from knowledge of his attitudes about the object. 74 In this respect, Rokeach's (1968) conceptualization does not differ greatly from that of many other theorists. The problem with these dispositional formulations is that rarely has it been possible to predict, with any substantial de- gree of accuracy, actual behavior in real-life situations solely on the basis of attitude test scores. Although agreeing with many other theorists in emphasizing that an attitude leads to a preferential (or discriminatory) response, there is an important difference between Rokeach's (1968) dispositional formulation and others. Since an attitude is conceptualized as an organ- ization of beliefs, it "is not a single predisposition but a set of interrelated predispositions focused on an atti— tude object or situation" (1968, p. 453). Each of the be- liefs comprising an attitude represent predispositions which lead to a preferential response when activated, and all attitudes are thus assumed to have a behavioral com- ponent. All of these predispositions, however, will not necessarily be activated by a particular attitude object or situation. "Which ones are activated depends on the particular situation within which a particular attitude object is encountered" (1968, p. 453). Rokeach's (1968) conceptualization of attitude leads, then, not only to the hypothesis that behavior can be predicted from attitudes, but to the additional hy- pothesis that in order to predict accurately, at least two attitudes, A0 and AS, as well as their cognitive 75 interaction, must be considered. Results of the present study support both of these hypotheses. The finding that the correlations between A0 and B are statistically signifi- cant in only two of seven instances and that the corre- lations between AS and B are statistically significant in only three of seven instances are not atypical of results from studies in which only one class or type of attitude is utilized to make predictions about behavior. Use of Rokeach's conceptualization, however, implies that at least two attitudes and their cognitive interaction are required to predict accurately. In the present study, the correlations between the AOAs configurations and B are statistically significant in six of the seven subsamples. The conceptualization of attitude from which the hypotheses of the present study were derived, and the re- sults of the present study, have some important impli— cations for both previous and for future research on the relationships between attitudes and behavior. Some Implications for Previous and for Future Research Some of the implications of Rokeach's (1968) con- ceptualization of attitude for the interpretation of the results of previous studies of the relationships between attitudes and behavior have been pointed out in a previous section. The studies by LaPiere (1934) and by Kutner EE.El° (1952) are those most often cited as examples of the inconsistency which is often found between expressed 76 attitudes and overt behavior. The results of these two studies, and others of a similar nature, have led some investigators to minimize the efficacy of verbal attitudes as accurate predictors of behavior. Recognizing the importance of the situation within which an attitude object is encountered, some theorists have suggested that assessment of the situational vari- ables will lead to more accurate prediction than will assessment of attitudinal variables. This view has re- sulted in the proposal of constructs such as "situational thresholds" (Campbell, 1963), "social involvement" (DeFleur and Westie, 1958), and "commitment" (Fendrich, 1967), to explain away the inconsistencies which are found between expressed attitudes and overt behavior. Rokeach's (1968) discussion of the nature of atti— tudes, however, proposes that these theorists have failed to appreciate that attitudes may be toward situations as well as toward objects, and may be assessed by similar means. Since an attitude object is always encountered within some situation about which an individual also has an organization of beliefs, prediction of behavior must be based upon consideration of both kinds of attitudes. Knowledge of these two types of attitudes alone, however, is not sufficient. The two kinds of attitudes will cog— nitively interact with one another. Behavior thus depends upon the outcome of this cognitive interaction. The same object may be encountered in two differently evaluated 77 situations, the relative importance of the situation being greater in one instance than the other, thus resulting in different behavior with respect to the object. In the present study, the cognitive interaction between A0 and As was assessed by means of an adaptation of the Rokeach and Rothman (1965) belief congruence model. The correlations between the AOAS configuration and B are statistically significant in six of seven instances, and are higher than the correlations between either A0 and B or A8 and B in every one of the seven subsamples. These results therefore support the prOposition that assessment of the two types of attitudes and their cognitive inter- action leads to more accurate prediction of behavior than does assessment of only one type of attitude. This sug- gests that LaPiere (1934) and Kutner EE.§l° (1952) may have found different results if evaluations of the relative im- portance of the two situations (face-to-face and non-face- to-face) had been assessed in addition to evaluations of the attitude objects. Analogous suggestions may be offered in regard to studies of a similar nature. It should be noted at this point that Rokeach (1968) is by no means minimizing the importance of the situation within which an attitude object is encountered. Rather than hypothesizing about situational constructs, however, it is prOposed that these situational constructs be psychologically reformulated as "attitude toward situ— ation." 'Attitude toward situation may be, and, it is 78 implied, often is relatively more important than attitude toward object. In the present study, for example, the mean evaluation and the mean perceived relative importance of the situation is greater than that of the attitude ob- jects in all seven of the subsamples. The major implication of Rokeach's conceptualization of attitude, and of the results of the present study, for future investigations of the relationships between atti— tudes and behavior, then, is that both A0 and AS, as well as their cognitive interaction, must be assessed in order to make accurate predictions about behavior. The investi— gators' task is to identify the relevant A0 and AS when making predictions about a specific behavior. This identi- fication of the relevant attitudes will often be a matter of intuition or of educated guesses based upon previous research. In the present study, for instance, it was assumed that in predicting class cutting behavior the most relevant attitudes would be those toward the instructor of a particular course (A0) and toward the general importance of attending class (AS). In addition to identifying the relevant attitudes, it is also necessary to assess the absolute and the rela— tive importance of these two attitudes in order to predict the behavioral outcome of their cognitive interaction. The question may be raised as how to assess the relative importance of the two attitudes. One way is exemplified by the present study in which Ss provided measures of (w) 79 and (1-w) by rating the relative importance of A0 and As in the context of AOAS. It is also possible, however, to assess the relative importance of the two attitudes by theoretical rather than empirical means. Rokeach (1966-67) has suggested that the compari- son of the relative importance of the two attitudes will take place within the general framework of one's total belief system. According to some tentative conceptuali- zations of the structure of belief systems (Rokeach, 1960, 1963, 1964), all beliefs and thus all attitudes, which are organization of beliefs, are arranged along a central- peripheral dimension of importance. Rokeach's conceptuali- zations of the types of beliefs and attitudes which may be found along this dimension of importance allow at least some educated guesses to be made about which of the two attitudes, A0 or AS, is likely to be more important. Be- cause of the composition of the sample in the present study, for example, it was hypothesized that AS would be relatively more important than A0. The empirical rating of (w) and (l-w) support this hypothesis. In regard to the relative importance of A0 and AS, it might be suggested that AS will often be the more im- portant of the two. Findings discussed in previous sections, those of Biesanz and Smith (1951), LaPiere (1934), Lohman and Reitzes (1954), Minard (1952), and Kutner g£_§l. (1952), for example, all emphasize the importance of the situation within which an attitude object 80 is encountered. Findings of the present study also show the greater relative importance of AS than of A0. The implication for future studies of the relationships be- tween attitudes and behavior is clear. The importance of situational attitudes and their interactions with attitudes toward objects have been too often overlooked in the past. Limitations 2: Present Study The proposition that behavior is a result of at least two cognitively interacting attitudes implies that exact prediction of behavior is possible, provided that the proper measurements are made and that these measure- ments have a high degree of reliability and validity. In the present study, the range of correlations between the AoAs configuration and B is from -.38 (subsample B5) to -.78 (subsample B6). The amount of variability in B accounted for by knowledge of the cognitive interaction between A0 and As thus ranges from 14 to 61%. Some of the limitations of the present study suggest that more accurate prediction is possible in future studies which attempt to overcome these limitations. Perhaps the most obvious limitation is that be- havior was not actually observed. The Ss were asked to estimate the number of class cuts for each course. As previously noted, however, there was no reason to assume that these estimates would not be reasonably accurate or that the Ss would be dishonest in making them. It would 81 seem.desirable, though, to have actual observations of behavior in future research, when time and conditions permit, or at least to have checks on the reliability of ‘Ss' reports of behavior. Such checks need not necessarily be made on the entire sample. A properly selected sub- sample could conceivably provide estimates of any dis- crepancies between actual behavior and verbal reports of such behavior. Related to the measurement of responses is the possibility that in the present study a S's report of his behavior and his evaluations of the objects and situation may not have been entirely independent of one another. It is conceivable, for instance, that if a S had indicated a large number of cuts in a particular course, he would, for reasons of consistency or rationalization alone, indi- cate a low evaluation of the instructor of that course. If such an effect does exist, it might be minimized by randomization of questionnaire items or by obtaining the different measurements at diverse and unrelated points in time. Two additional aspects of the behavioral measures should be noted. Although the Ss were instructed to dis- regard factors such as illness, accident, and unusual weather conditions when estimating the number of times they cut a particular course, there is no way of knowing to what extent these instructions were followed. The Correlations between the attitudinal and behavioral 82 measures, as well as the reliability and validity of the behavioral measures, would be attenuated to the extent that these instructions were not kept in mind when esti— mating B. One way of discovering if instructions were followed would have been by interviewing the Ss. Another aspect of the behavioral measure worth consideration is the question of the linearity of class cutting behavior. That is, the question may be raised as to whether, say, two cuts are psychologically twice as many as one cut, and whether the difference between one cut and two cuts has the same meaning as that between three cuts and two. This latter question requires further investigation and should be kept in mind in future studies. Implicit in the formulation that behavior is a function of at least two cognitively interacting atti- tudes is the proposition that the prediction of future behavior is based upon prior knowledge of A0 and As and of their relative importance with respect to one another. In the present study, Ss were asked for their evaluations of the objects and of the situation after the actual be- havior had taken place. No measures of possible attitude change over time were obtained. It does not seem un- reasonable that such changes in evaluation, especially of the course instructors, would occur. Ideally, a test of the formulation represented by B = f(AO,AS) requires that the two attitudes by relatively stable, or at least that they be assessed as close in time 83 as possible to the occurrence of the behavior. Since it will not always be possible to meet these ideal conditions, investigators must be aware of what the A0 and As measures actually represent. In the present study, for example, the A0 and AS measures may be viewed as mean or average evaluations of the objects and the situation over the period of time in which the S3 encountered the particular objects within the particular situation. The possibility of a difference among Ss in the meaning of the A0 and AS measures should also be noted. In evaluating the course instructors, Ss were asked to what extent they "liked" each particular instructor. "Liking" can be interpreted in a number of different ways. To some Ss it may have meant "good lecturer," to others "a nice person," and so on. Differences in interpretation of "the general importance of attending class" are also possible. These considerations illustrate the importance of clear and explicit instructions and the desirability of interviewing Ss in order to increase the reliability and validity of obtained measurements. The small variability and the skewness exhibited by some of the obtained distributions must also be kept in mind when attempting to interpret the results of the present study. The distributions of B and of AS in par- ticular, are limited in range and show considerable skew- ness. How does this affect the interpretation of the obtained £_values? 84 In regard to the small ranges, it will be recalled that the general effect of increasing the variability of a distribution of measurements is to increase the values of the correlations. Decreasing the range of measurements, conversely, tends to decrease the size of the correlations. It should be expected, then, that the £_values obtained in the present study would be even larger if the distributions had shown more variability. Since extremely skewed distri- butions exhibit less variability than unskewed distribu- tions, the same reasoning applies; the obtained Es would be expected to be larger if the distributions had been less skewed. Another factor which may have attenuated the values of the obtained correlations is that no corrections for possible overassimilation effects were made. Reasons for not making corrections for possible overassimilation effects are discussed above, where it is pointed out that is overassimilation does occur it will work against the hypothesis that AOAS is a better predictor of B than is either A0 or AS alone. The possibility of such over— assimilation effects, however, must be kept in mind in future investigations. Transformations of the data might have been per- formed before analyzing the results of the present study. The distributions showing considerable skewness and little variability, for example, might have been transformed into a normal form. The effect of doing so would probably have 85 been to increase the correlational values and the level of their significance. Such transformations were not per- formed for two reasons. One justification for transform- ing obtained distributions into a normal form is an assump- tion that the variables are in fact normally distributed in a population, even though those of a particular sample are not. Since the population of interest in this study is that of college students, there seems to be little justification for the assumption that the obtained measures, As and B in particular, are normally distributed. A second reason for not performing transformations on the data is that when transformations are made, any discussions of, or generalizations from, the analyses can, strictly speak- ing, be only about the transformed measures and not about the variables which they represent. Other aspects of the present study, such as the homogeneity of the sample, the sample size, and the par- ticular attitudes and behavior studied, must also be kept in mind when attempting to generalize from the results. Even though results of this study support the proposition that behavior depends upon at least two cognitively inter- acting attitudes, further investigations need to be carried out on samples from diverse types of populations and with various types of attitudes and behaviors. To summarize this section on limitations of the present study, it should be noted that some aspects of the study may have affected the reliability and validity 86 of the measurements. These include the following facts: (a) the behavioral measures represent estimates provided by the Ss after the occurrence of the actual behavior, (b) attitude change may have occurred, (c) factors other than A0 and As may have affected class cutting behavior, and (d) the meaning of the behavioral and attitudinal measures may have differed among the Ss. Some factors which may have tended to limit the values of the obtained corre- lations include the facts that: (a) the obtained distri- butions show considerable skewness and little variability, and (b) no corrections for possible overassimilation effects were made. In conclusion, then, results of the present study lend support to the general hypothesis that behavior can be predicted from knowledge of verbally expressed atti- tudes. The more specific hypothesis that prediction of behavior with respect to an object within a situation will be more accurate when based upon knowledge of the cogni- tive interaction between A0 and As than when based upon knowledge of either attitude alone is also supported by the results. Evidence for the greater relative importance of As than of AC was also found, although this result would be expected to vary with the particular objects and situations chosen for study. REFERENCES Adorno, T. W., Frenkel-Brunswik, E., Levinson, D. J., & Sanford, R. N. The authoritarian personality. New York: Harper, I950. Allport, F. H. Social psychology. Boston: Houghton Mifflin, 1924. Allport, G. W. Attitudes. In C. Murchison (Ed.), Hand- book 2: social 5 cholo . Worcester, Mass.: Clark Univer. Press, I935. Pp. 798—844. Biesanz, J., & Smith, L. M. Race relations of Panama and the Canal Zone. American Journal 9E Sociology, 1951, 57, 7-14. Blumer, H. Attitudes and the social act. Social Problems, 1955, 3’ 59-640 Blumer, H. Research on race relations in the United States of America. International Social Science Bogardus, E. 8. Fundamentals 9: social 8 cholo . (2nd ed.), New York: Century, 193%. Boneau, C. A. The effects of violations of assumptions underlying the 5 test. Psychological Bulletin, 1960, 57, 49-64. Brehm, J. W., & Cohen, A. R. Explorations i3 cognitive dissonance. New York: Wiley, I962. Campbell, A., Converse, P. E., Miller, W. E., & Stokes, D. E. The American voter. New York: Wiley, 1960. Campbell, D. T. The indirect assessment of social V/ attitudes. Psychological Bulletin, 1950, 47, 15-38. 87 88 Campbell, D. T. Social attitudes and other acquired be- havioral dispositions. In S. Koch (Ed.), Ps - cholo : A stud of a science. Vol. 6. New York: Mch-aw-Hi 1,—1963. Pp. 94-172. Chein, I. Behavior theory and the behavior of attitudes: Some critical comments. Psychological Review, 1948, 55, 175—188. Cook, 8. W., & Sellitz, C. A multiple—indicator approach to attitude measurement. Psychological Bulletin, 1964, 62, 36-55. DeFleur, M., & Westie, F. Verbal attitudes and overt acts: An experiment on the salience of attitudes. American Sociological Review, 1958, 23, 667-673. DeFleur, m., & Westie, F. Attitude as a scientific con- cept. Social Forces, 1963, 42, 17—31. Doob, L. W. The behavior of attitudes. Psychological Review, 1947, 54, 135-146. Dreyer, E., & Rosenbaum, W. A. (Eds.), Political gpinion and electoral behavior: Essays and—Studies. Belmont, Calif.: Wadsworth, 1966. Fendrich, J. A study of the association among verbal attitudes, commitment and overt behavior in different experimental situations. Social Forces, 1967, 45, 347—355. Festinger, L. A theory 2; cognitive dissonance. Evan- ston, IIl.: Row, Peterson, 1957. Fishbein, M. A consideration of beliefs, attitudes, and their relationship. In I. D. Steiner & M. Fish- bein (Eds.), Current studies i3 social psychology. New York: Holt, I965. Pp. 107-120. Fishbein, M. Attitude and the prediction of behavior. In M. Fishbein (Ed.), Readin s in attitude theory and measurement. New York: WiIEy, 1967. Pp. 47'7-492. Fishbein, M. (Ed.), Readings in attitude theor and measurement. New York?_‘WiIey, I967. Fishman, J. Some social and psychological determinants of inter-group relations in changing neighborhoods: An introduction to the Bridgeview study. Social Forces, 1961, 40, 43-51. 89 Folsom, J. K. Social psychology. New York: Harper, 1931. Fromm, E. Esca e from freedom. New York: Farrar and Rine art, I9II. Gordon, R. L. Attitude and the definition of the situation. American Sociological Review, 1952,_l7, 50—58. Green, B. F. Attitude measurement. In G. Lindzey (Ed.), Handbook 9: social psychology. Vol. 1. Theor and method. Cambridge, Mass.: Addison-Wesley, I954. Pp. 395-369. Harding, J., Kutner, B., Proshansky, H., & Chein, I. Prejudice and ethnic relations. In G. Lindzey (Ed.), Handbook 2: social psychology. Reading, Mass.: Addison-Wesley, 195 . Pp. 1021-1061. Hays, W. L. Statistics for psychologists. New York: Holt, 1963. Horowitz, E. L. "Race" attitudes. In G. Klineberg (Ed.), Characteristics 9: the American Negro. New York: Harper, 1944. Pp. 139-247. Insko, C. Theories g: attitude change. New York: AppIeton-Century, 1967. Jahoda, M., & Warren, N. (Eds.), Attitudes. Baltimore: Penguin Books, 1966. Janis, I. L., & King, B. T. The influence of role-playing on Opinion change. Journal of Abnormal and Social Psychology, 1954, 49, 2II-2187 Katz, D., & Stotland, E. A. A preliminary statement to a theory of attitude structure and change. In S. Koch (Ed.), Psychology: A study of a science. Vol. 3. Formulations of Ehe ersan and the social context. New York: MEEraw:HilI, 1959T'_95} 423- King, B. T., & Janis, I. L. Comparison of the effective- ness of improvised versus non-improvised role playing in producing Opinion changes. Human Relations, 1956, 9, 117-186. Krech, D., Crutchfield, R., & Ballachey, E. L. Individual £3 society. New York: McGraw-Hill, 1962. Krueger, E. T., & Reckless, W. C. Social psychology. New York: Longmans, Green, 1931. 90 Kutner, B., Wilkens, C., & Yarrow, P. R. Verbal attitudes 7’ and overt behavior involving racial prejudice. Journal of Abnormal and Soc1al Psychology, 1952, 47, 649-652. LaPiere, R. T. Attitudes vs. actions. Social Forces, 1934, 13, 230-237. Lewin, K. Field theory and experiment in social psy- chology: Concepts and methods. American Journal p£_Sociology, 1939, 44, 868-897. Lewin, K. Behavior and develOpment as a function of the total situation. In D. Cartwright (Ed.), Field theory in social science: Selected theoret1caI papers 5_ Kurt Lewin. New York: Harper, 1951. Pp. 8- 03. Linn, L. Verbal attitudes and overt behavior: A study of racial discrimination. Social Forces, 1965, 43, 353-364. Lohman, J. D., & Reitzes, D. C. Deliberately organized groups and racial behavior. American Sociological Review, 1954, 19, 342-348. Minard, R. D. Race relationships in the Pocahontas coal field. Journal 9; Social Issues, 1952, 8, 29-44. Murphy, G., Murphy, L. B., & Newcomb, T. M. Experimental social psychology. New York: Harper, 1987] Orne, M. T. On the social psychology of the psychological experiment, with particular reference to demand characteristics and their implications. American Psychologist, 1962, 17, 776—783. Osgood, C. E., & Tannenbaum, P. H. The principle of con- gruity in the prediction of attitude change. Psychological Review, 1955, 62, 42-55. Pettigrew, T. F. Social psychology and desegregation research. American Psychologist, 1961, 16, 105-112. Raab, E., & Lipset, S. M. The prejudiced society. In E. Raab (Ed.), American race relations today. New York: Doubleday, I962. Pp. 29-55. Rokeach, M. The open and closed mind: Investigations into the nature pg belief systems and personality systems. New York: Basic Books, 1960. 91 Rokeach, M. The organization and modification of beliefs. Centennial Review, 1963, 7, 375—395. Rokeach, M. The three Christs g: Ypsilanti: A_psycho- logical study. New York: Knopf, 1964. Rokeach, M. Attitude change and behavioral change. w/ Public Qpinion Quarterly, 1966-67, 30, 529-550. IRokeach, M. Attitudes. In D. L. Sills (Ed.), Inter- national encyclopedia 9: the social sciences. Vol. I. New York: Macmillan, 1968. Pp. 449—458. Rokeach, M., & Rothman, G. The principle of belief con- gruence and the congruity principle as models of cognitive interaction. Psychological Review, 1965, 72, 128-142. Rose, A. M. Intergroup relations vs. prejudice: Pertinent theory for the study of social change. Social Problems, 1956, 4, 173-176. Rotter, J. B. The role of the psychological situation in determining the direction of human behavior. In M. R. Jones (Ed.), The Nebraska Symposium. Univer. of Nebraska Press, I955. Saegner, G., & Gilbert, E. Customer reactions to the integration of Negro sales personnel. International Journal of Qpinion and Attitude Research, 1950, 4, '573767—'_' Scott, W. A. Attitude change through reward of verbal ,2 behavior. Journal gf_Abnorma1 and Social Psy- Scott, W. A. Attitude change by response reinforcement: Replication and extension. Sociometry, 1959, 22, 328-335. Sherif, M., & Cantril, H. The psychology 2: ego-involve- ments. New York: Wiley, 19 . Shibutani, T. Society and ersonalit . Englewood Cliffs, New Jersey: Prentice-Hall, I961. Smith, M- B., Bruner, J. S., & White, R. W. Opinions and personality. New York: Wiley, 1956. Thomas, W. I., & Znaniecki, F. The Polish peasant ip Europe and America. Chicago: Univer. Chicago Press, I9I8. 92 Weissberg, N. On De Fleur and Westie's "Attitude as a scientific concept." Social Forces, 1965, 43, 422-425. FEB 261989 I11111111111s