THE METATHEORY OF FACETS: CONSTRUCT VALIDITY ' or A STRUCTURAL APPROACH TO ATTITUDE MEASURMENT Dissertation for the Degree of Ph. D. MICHIGAN STATE UNIVERSITY STEPHEN KENT BEDWEIJ. 1977, 1&5 W, _._-- I II III IIIIIIIIII I II I III! III III II I; w L a ‘. . - (a ' k... up "(REC University ’ 1" - . 1",-J;-,{.l.._\lll-.E I r This is to certify that the thesis entitled The Metatheory of Facets: Construct Validity ofA Structural Approach to Attitude Measurement presented by Stephen K. Bedwell has been accepted towards fulfillment of the requirements for Ph.D. degree in Counseling, Personnel Service & Educational Psychology \ 7 {Ir/fi'é: C2} 1%??? L ‘77 / Wrofessor Date November 22, 1976 0-7639 [Tl ",lo\’,/ ’9": 1 I! .’ ABSTRACT THE METATHEORY 0F FACETS: CONSTRUCT VALIDITY OF A STRUCTURAL APPROACH TO ATTITUDE MEASUREMENT By Stephen Kent Bedwell The present study examined the construct validity of facet theory applied to attitude measurement. Earlier research has used one of two approaches in defining attitude: either emphasizing atti- tude as a "predisposition" to behavior or regarding attitude as "behavior" per se. Recently the affective-cognitive-conative notion of attitude has been held by a majority of attitude theorists. Jor- dan (1968) has concluded that most research studies have been incon- clusive or contradictory about attitudes because attitude scales utilized were composed of items stemming from different structures: i.e., from different levels of attitude (for example, the cognitive- affective-conative) of the universe of attitudes toward specified objects. Guttman (l950) has operationally defined attitude as "a delimited totality of behavior with respect to something." Jordan and Guttman have stated that it is productive to drop the dichotomy between attitude and behavior and have the term "attitude" embrace both varieties, the predisposition to respond and the response itself with "subvarities ranging from stereotypic generalizations to overt instrumental behaviors . . . ." Thus the term "attitude-behavior." Stephen Kent Bedwell The purpose of the present research was to further examine relation- ships between the cognitive-affective-conative components of atti- tude across the attitude-behavior levels toward various attitude objects. The metatheory of facets specifies certain structural out- comes of correlations dependent upon specific roles played by the facets (which have been incorporated in the design) and the structue ples (a combination of elements of each facet) as the structuples become increasingly stronger. Various researchers have investigated and obtained the "simplex" (a structural outcome) relation predicted by facet theory and thus support for the construct validity of the theory has been obtained. No attitude studies have investigated the construct validity of facet theory where the joint (attitude levels) structioned and lateral (situations) structioned facets have been held constant across selected attitude objects. In brief, the methodology of this study involved three attitude-behavior scales utilizing facet procedures toward physical, mental, and social attitude objects: (a) a mental retardation scale, (b) a race scale, and (c) a blind scale. These attitude objects were placed in the "same" social distance life situations (lateral structioned) at three joint structioned levels: stereotypic, moral evaluation, and personal feeling. The research was designed to con- trol for sources of variation due to the joint and lateral dimensions, by holding constant these dimensions and changing only the subject- object in situation relationship. The scales were administered to a homogeneous sample to enable reduction of variance dueix>differential Stephen Kent Bedwell contact with the attitude objects, social class, age, etc. The order of scale administration was balanced to control for progressive error and response set (Underwood, 1966). Five research hypotheses were stated and, with exception of the order of scale administration effecting attitudes toward specific attitude objects, the results supported the hypothesis. A simplex relation was obtained for each of the social distance (lateral struc- tioned) scales at each joint structioned level for each attitude object. The simplex structure was obtained for each facet-derived subscale at each level for each object. The cylindrex structural hypothesis was approximated: where joint structioned facets served an axial role, social distance situations served a modulating role, and attitude object served a polarizing role. And finally, smallest space analysis resulted in fewer dimensions than did factor analysis. It was concluded that the construct validity of the meta- theory of facets was further established, and that facet theory is a useful tool for specifying research designs, and in a priori struc- turing the relationships. It was recommended that further studies investigating an ordering principle for attitude objects, clarifying the lateral structioned (situations) facet to establish a more clear ordering principle and expanding upon and using a more heterogeneous population, are necessary. Without exploiting the data, it was fur- ther concluded that facet theory may be an extremely useful tool in designing classical experimental research designs, and perhaps even- tually specifying counselor therapeutic relationships. THE METATHEORY OF FACETS: CONSTRUCT VALIDITY OF A STRUCTURAL APPROACH TO ATTITUDE MEASUREMENT By Stephen Kent Bedwell A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Counseling, Personnel Services, and Educational Psychology 1977 To my wife, Bernadette A. Bedwell, and her brother, Senator Anthony A. Derezinski: the two people who most influenced my educa- tional pursuits. And to my parents, Clyde E. and Norma L. Bedwell. ii ACKNOWLEDGMENTS First, I would like to express my thanks to my advisor, chairman, and personal friend, Dr. John E. Jordan, for his unfailing confidence in my abilities, his guidance and encouragement in all aspects of this dissertation. I also would like to express my appreciation to my committee members, Dr. Thomas Gunnings, Dr. James Engelkes, and Dr. Ronald Nolthuis, for serving as members of my doctoral committee. Also, my thanks to Dr. Ward Wilson for assisting in the collection of the data. Special thanks goes to Dr. James Lingoes for offering valuable assistance in the use of Smallest Space Analysis and the Guttman- Lingoes Nonmetric Series at the University of Michigan. Finally, to Professor Louis Guttman, I offer my sincere appreciation for explaining aspects of facet theory and nonmetric techniques, and his valuable critique of this study. Furthermore, without the valued support of my friend Dr. James Stratoudakis who listened to my frustrations and offered his advice, this project would have been overwhelming. Muchas gracias to Guadalupe Solis for typing the final draft of this dissertation. To my wife Bernadette goes the role of supporter, pusher, and confidante: she willingly typed many drafts of this thesis; she gave of her valued and limited relaxation time to listen to my iii ideas and frustrations. Without her support this dissertation and Ph.D. would not have been possible. Also, bless my children who gave of their "Daddy time"; Natalie, Anthony, and Vincent, Daddy loves you. iv TABLE OF CONTENTS LIST OF TABLES LIST OF FIGURES . PREFACE Chapter I. INTRODUCTION Overview: Definitional Status of Attitude Constructs . Facet Theory Approach to Attitude Measurement Purpose . . . . . . . . . II. FACET DESIGN, STRUCTURAL THEORY, AND ATTITUDE MEASUREMENT . . . Guttman' 5 Post Hoc Facetization of an Attitude Research Problem . . . . Structural Relations Specified by Facet Procedures . . . Nonmetric Analysis: Examination of Structural Order and Lawfulness . . Facet Theory and Attitude Measurement . Summary of Findings of Previous ABS Scales. Validity of Facet Theory Applied to Attitude Measurement III. DESIGN AND METHODOLOGY . Overview of Methodology. . . . . Ordering Principle of Facet Design: Lateral Struction . . . . . . Selection of a Homogeneous Sample Scale Administration Procedures Statistical Analysis . . Preparation of Data . . . Multivariate Analysis of Variance . Simplex Approximation Test Smallest Space Analysis I Factor Analysis . . Page vii ix xi (DUI—l I3 17 19 ‘ 27 31 Chapter Hypotheses . . . . . . . . . Hypothesis l: Scale Administration Hypothesis 2: Simplex for Social DistanCe : Hypothesis 3: Simplex for Facetized Scales Hypothesis 4: Structure of Facet Derived Scales Hypothesis 5:. Factor Analysis Compared to . Smallest Space Analysis IV. RESULTS . Description of Sample . . Scale Scores and Reliability Examination of Hypotheses . Hypothesis l: Scale Administration Hypothesis 2: Simplex for Social DistanCe : Hypothesis 3: Simplex for Facetized Scales Hypothesis 4: Structure of Facet Derived Scales . . . . . . . . . . . Hypothesis 5: Factor Analyis Compared to Smallest Space Analysis . . . V. DISCUSSION . Overview of Purpose and Methodology . Order Effect . Simplex for Social Distance Items . Smallest Space Analysis and the Structural Hypotheses . . Factor Analysis . . Recommendations for Future Research. APPENDICES A. GLOSSARY B. ATTITUDE-BEHAVIOR SCALE: ABS-BK, MR, 80 . REFERENCES vi Page 100 103 104 106 107 III 135 Table 10. ll. 12. l3. )4. LIST OF TABLES Simplex Structure of Correlation Matrix . Intercorrelations for an Equally Spaced, Uniform, Perfect, Additive Circumplex . . . Comparison of Guttman and Jordan Facet Designs for Attitude Items Permutations of Five Two-Element Facets of Table 3 Combinations of Five Two-Element Facets and Basis of Elimination Five-Facet Six-Level System of Attitude Verbalizations: Levels Facet Profiles, and Definitional Statements for Twelve Combinations . . Joint Level, Profile Composition, and Labels for Six Types of Attitude Struction . . . . Facets Used to Determine Joint Struction of an Attitude Universe . . . Order of Scale Administration Arithmetic Means for Joint Level Scores by Attitude Object and Demographic Classification Variables Correlation of Social Distance Items to Social Distance Subscore Total and Joint Level Total Scores for the Black Attitude Objects . . . Correlation of Social Distance Items to Social Distance Total and Joint Level Total Scores for the Mentally Retarded Attitude Objects . . . Correlation of Social Distance Items to Social Distance Subscores Total and Joint Level Total Scores for the Blind Attitude Objects . . . . Cell Means by Order of Scale Administration, Object, Joint Structioned Level, and Social Distance Subscores . . . . . . . . 'vii Page 22 23 34 36 37 38 39 50 58 65 67 68 69 7] Table 15. 16. 17. 18. 19. 20. 21. Multivariate Analysis of Variance for Order Effect by Object, Level and Social Distance . Simplex Matrices for Lateral Structioned Social Distance by Joint Structioned Attitude Level for Black Attitude Objects . . . Simplex Matrices for Lateral Structioned Social Distance by Joint Structioned Attitude Level for Mentally Retarded Attitude Objects Simplex Matrices for Lateral Structioned Social Distance by Joint Structioned Attitude Level for Blind Attitude Objects . . . Correlations for Social Distance Subscores by Joint Structioned Level and Attitude Objects . Guttman-Lingoes' Smallest Space Coordinates for a Three-Dimensional Space . . . . . Rotated Factor Matrix (Varimax) of 27 Depending Scores (Loadings .40) . . viii Page 73 75 76 77 79 81 94 LIST OF FIGURES Figure 1. 2. 3. 10. 11. 12. A Diagram of a Hypothetical Radex . A Hypothetical Cylindrex . A Mapping Sentence of the Joint, Lateral, and Response Mode Struction Facets Used to Structure the Attitude-Behavior Scale-Mental Retardation A Mapping Sentence for the Facet Analysis of Attitudes Toward Blacks, Mentally Retarded and Blind . A Three-Dimensional Representation of Three Joint Structioned Levels and Three Social Distance Sub- scores for Three Attitude Objects . Three-Dimensional Representation of Moral Evaluation Structioned Level and Social Distance Scores for Three Attitude Objects . . . . Three-Dimensional Representation of Stereotypic Structioned Level and Social Distance Scores for Three Attitude Objects . . . . . . . Three-Dimensional Representation of Personal Feeling Structioned Level and Social Distance Scores for Three Attitude Objects . A Three-Dimensional Representation of Joint Structioned Levels and Social Distance Subscores for "Blind" Attitude Objects . . . A Three-Dimensional Representation of Joint Structioned Levels and Social Distance Subscores for "Black" Attitude Objects . - . . . . A Three-Dimensional Representation of Joint Structioned Levels and Social Distance Subscores for "Mentally Retarded" Attitude Objects . . . Approximation of Obtained Cylindrex ix Page 24 26 41 55 82 83 84 85 86 87 88 89 Figure Page l3. A Two-Dimensional Plot for Moral Evaluation Axial Facet with Social Distance and Attitude Object . . . . . 90 l4. A Two-Dimensional Plot for Stereotypic Axial Facet with Social Distance, and Attitude Object . . . . 91 15. A Two-Dimensional Plot for Personal Feeling Axial Facet with Social Distance, and Attitude Object . . . . 92 PREFACE This study is an example of the project approach to gradu- ate research: it is one in a series, jointly designed by several investigators, each study acting as a building block to further extend and explore the theoretical underpinnings of facet theory. Therefore, similarities in the approach to research problems. the- oretical material, instrumentation, design, and analysis are both necessary and desirable. Nevertheless, some theoretical specifica- tions, localities, samples, necessary adaptations, and interpreta- tions in each study are those of the authors. xi CHAPTER I INTRODUCTION Overview: Definitional Status of Attitude Constructs It is axiomatic that attitudes are projected to have major import on decisions made in all areas of life, both for individuals and society. Daily, the media bombard the public with recent find- ings from various pollsters concerning the latest inclination of the public toward a particular attitude object: the economy, politicians, legislation, and Blacks--to name only a few. It is also projected that this activity may in fact generate attitudes in its own right. In addition, attitude is one of the most popular concepts in the social sciences, having been in use among psychologists for more than lOO years (Elizur, l970). This has been especially true since All- port's classic article (Allport, 1935) which gave emphasis and defi- . nitional status to the concept of attitude. Some social scientists have even suggested that social psychology be defined as the scien- tific study of attitude (McGuire, l969). However, there still exists no commonly accepted definition of the hypothetical construct of attitude. Two primary approaches have been used in defining attitude: one emphasizing attitude as a "predisposition" to behavior and the second regarding attitude as "behavior" per se. Behavior has been viewed as spanning the cognitive, affective, and conative domain of the human condition. More than 100 years ago, Herbert Spencer wrote about the attitude of the mind, referring primarily to the cognitive nature of attitude (Brodwin, l973; Elizur, l970). Most theorists use two cognitive elements in the definition of attitude: evaluation and beliefs. In many attitude studies (Elizur, 1970) respondents are frequently asked to rate objects. Osgood, Suchi and Tannenbaum (l957) developed the semantic differential technique as a method to measure the meaning of concepts and concluded that mean- ing is a location in a space defined by some number of factors or dimensions. An attitude toward a concept is its projection onto one of these dimensions defined as "evaluative." The semantic differ— ential researchers posit that attitude is expressed in terms such as good-bad, kind-cruel, honest-dishonest. Emotion regarding the attitude object is included in Thur- ston's definition: Attitude is the affect for or against a psycho- logical concept. ". . . Appetition is the positive form of affect, which in more sophisticated situations appears as liking, defending, or favoring. Aversion is the negative form of affect which is described as hating the object, disliking, or destroying it" (l931). Staats (l967) provides a more recent view which emphasizes the affective characteristics of attitudes. He defines attitude as an emotional response to a social stimulus, or a stimulus that has social significance (Fishbein, l967, p. 373). Recent attempts have been made to define attitudes in behavioral terms (Elizur, l970). Dobb (l947) suggests the following behavioral definition: "attitude is an implicit drive producing response, considered socially significant in the individual's soci- ety" (p. l35). A stronger behavioral definition is postulated by Green (l954) which attributes to attitude ". . . a consistency among responses to a specified set of stimuli or social objects“ (p. 335). This definition, according to Green, ". . . does divest attitudes of their affective and cognitive properties, which may be . . . cor- relates of the responses that comprise attitudes . . ." (p. 336). The systems approach (McGuire, l969) attempts to merge the cognitive, affective, and conative elements: ”As the individual develops, his cognitions, feelings, and action tendencies with respect to various objects in the world, become organized into enduring systems called attitudes . . J'(Elizur, l970, p. 37). Krech, Cruchfield and Ballachey (l962) utilize this approach and, in their system, the person's feelings, cognitions, and action ten- dencies "become mutually interdependent." That is, the individual's cognitions "are influenced by his feelings, and action tendencies toward" an object and therefore "a change in his cognitions about the object would tend to produce changes in his feelings and attitude tendencies toward it" (pp. l39-l40). They hold that attitudes are "enduring systems of positive or negative evaluations, emotional feelings, and pro or con action tendencies with respect to social objects" (1962, p. 139). However, overt action is not included in this definition of attitudes, but "the social actions of the indi- vidual reflect his attitude." They describe the cognitive component as consisting of beliefs about the object, the most important being "evaluative beliefs" (p. 140). However, because of the inter- relatedness of the three attitude components, few attitudes exist in isolation, "most of them form clusters with other attitudes." For instance, what has been called religionism is a cluster which accounts for attitudes toward evolution, God, and birth control. Some writers regard beliefs about an object as a measure of attitudes. Fishbein and Ravin (l962) suggest a different definition of belief as the probability dimension of the concept: for example, "is it existent or nonexistent, possible or impossible," etc. But of many objects the existence can hardly be doubted. Therefore, Fishbein and Ravin posit that belief about the object must be included (Elizur, l970). Rokeach (1968) provides a distinction between value and belief: a value is a single global belief, trans- cending object situation specificity and serving as a standard for judging, acting, valuing, and comparing. Rokeach has defined atti- tude as a package of beliefs, each of which is object and situation specific, serving as predispositions to act. An attitude is defined as "a relatively enduring organization of beliefs around an object or situation predisposing one to respond in some preferential manner" (Rokeach, 1968, p. ll2). In summary, most of the conceptions of attitudes are multi— dimensional in character: the affective-cognitive-conative notion is held by perhaps a majority of attitude theorists. The notion in brief is that an attitude is a somewhat enduring system of (a) beliefs, especially evaluative beliefs; (b) positive or nega- tive affect directed toward the object of the attitude and related objects; and (c) action tendencies regarding the object and its related objects. Many definitions of attitude, therefore, include the princi- ple that it is a predisposition to act in a certain way or an action tendency. The view of attitude as something which helps to predict a specific overt behavior is criticized by many researchers and the- orists. Guttman (l950) states directly that while some variables are predictive of an attitude, they form no part of the attitude. For example, while level of education may help predict a person's resistance to change, it is no part of the definition of attitude toward change (Elizur, l970). According to this argument, behavior depends not only on the attitude, but also on the situation (Guttman, l950, pp. 50-5l). Facet Theory Approach to Attitude Measurement In reviewing the literature, Jordan (l968) concluded that four classes of variables are important determinants, correlates, and/or predictors of attitudes: (a) demographic factors, such as age, sex and income; (b) sociopsychological factors, such as a per- son's value structure; (c) contact with the object and enjoyment of that contact; and (d) knowledge about the attitude object. In the review, Jordan (l968) concluded that most research studies were inconclusive or contradictory about the predictor variables and suggested that one reason might be that the attitude scales were composed of items stemming from different structures, i.e., from different levels of attitude (for example, the cognitive, affective, or conative) of the universe of attitudes toward specified objects. The lack of control over which attitudinal "levels" were being mea- sured was projected by Jordan to produce inconsistent, centradictory, and noncomparable findings in attitude research. Guttman (1953, 1959) developed the use of facet theory to analyze attitude. Facet theory is somewhat similar to factor analy- sis as used by Thurston and Spearman (Brodwin, 1973) but distinct in its a priori nature. Guttman (1950) operationally defined attitude as "a delimited totality of behavior with respect to something" (p. 51). Guttman's definition of attitude is consistent with the traditional distinction made between attitude and behavior, the difference between the inclination to act and the act itself (Jordan, 1971). Guttman's definition is also consistent with the previous distinctions, since attitude items are considered verbalizations of predispositions. Jordan (1971) used Guttman's definition to link attitude and behavior in the development of his facetized attitude- behavior scales. Jordan and Guttman (1976) state that it is productive to drop the traditional dichotomy between attitude and behavior and have the term "attitude" embrace both varieties, the predisposition to respond and the response itself, with "subvarieties ranging from stereotypic generalizations to overt and instrumental behaviors which are unfavorable to favorable to any personal or conceptual object" (p. 2). Initially, Guttman, using a facet theory approach, analyzed the work of Bastide and van den Berghe (1957) and posed four levels of an attitude universe: (a) stereotypic; (b) norm; (c) hypo- thetical interaction; and (d) personal interaction. Jordan (1968) expanded Guttman's four levels to six levels by adding the moral evaluation and personal action dimensions. According to Jordan (1971), Guttman's definition of attitude approximates the "positivistic definition" developed by McGuire (1969, p. 145) and facilitates a cognitive-affective-conative (know- ing, feeling, and acting) analysis of the human behavior. According to Jordan, Guttman's definition is consonant with a structural (Foa, 1965, 1968; Foa and Turner, 1970) approach to the facet analysis of attitude-behavior (Jordan, 1971, p. 7). Jordan postulates that ". . . if one is to do research on attitudes that is both socially relevant and methodologically rigorous . . .," Guttman's structural facet theory approach should provide (a) a definition of the research problem; (b) the selection of variables for study; and (c) the structuring of the relationship between the dependent and indepen- dent variables. Consequently, Jordan and his associates have taken a step toward merging the concept of attitude as a "predisposition" to behavior, to include behavior itself. His concept of attitude- behavior and the six attitudinal levels facilitates an examination of the relationship between the cognitive-affectiVe-conative com- ponents as well as emphasizes the conative component as the cri- terion of behavior. Purpose The purpose of the present research was to further examine relationships between the cognitive-affective-conative components of attitude across the attitude-behavior levels toward various atti- tude objects in situation. In this research the use of Guttman's behavioral definition in the context of a more detailed facet theory analysis is presented. The generality of definition is justified by the "first law of attitude" and the growing number of specific empirical laws to which it leads (Gratch, 1973; Jordan, 1971). The first law asserts: If any two items are selected from the university of atti- tude items toward a given object, and if the object observed is not selected artificially, then the population regres- sions between these two items will be monotone, and with positive or zero sign (Gratch, 1973, p. 36). This law covers the cognitiveéaffective-instrumental behavior and thus shows the fruitfulness of considering all three variations as attitudinal (Jordan, 1971). Guttman (1950) suggests two basic premises for the defini- tion of a scientific concept: "(a) it must be defined in terms of observation; and (b) a definition is scientifically useful only insofar as it leads to objective research" (p. 49). Attitude in this research is regarded as a subclass of behavior and is thus con- sonant with Guttman's definition of attitude. Thus, attitude is regarded herein as the totality of behavior for or against an object, i.e., it can be observed in a degree of favorableness or unfavorableness of behavior toward the object. The term "attitude object" is used in its widest sense; it may refer to a physical, social, psychological, various situations and ideas, i.e., every- thing toward which a person can behave positively or negatively (Elizur, 1970). Regarding an attitude as a totality of behavior "toward an object" allows the term attitude-behavior toward an object to be used interchangeably. In this definition, overt and covert behaviors are part of an attitude, but at different levels, and cover cognitive, affective, and instrumental behaviors (Elizur, l970). Guttman's behavioral definition has been enlarged on in the context of the more detailed facet theory analysis developed by Jor- dan and his colleagues at Michigan State University. The present study is part of a series, all of which have aimed at developing further the facet theory attitude-behavior methodology. Since the inception of Jordan's project approach to graduate research, Attitude-Behavior Scales (ABS) have been developed to assess atti- tudes toward many "personal" attitude objects, such as Whites toward Blacks, Blacks toward Whites, mentally ill or emotionally disturbed persons, the deaf, the undereducated adult, the blind, the war dis- abled (in Vietnam), and drug users. Recently, attitude-behavior scales toward "conceptual" objects have been developed: the environ- ment, role of women, technical education, educational change, and affective education (Jordan, 1976). The validity of the ABS series of scales, using the known group method of establishing concurrent validity, has been successful (Brodwin, 1973; Jordan, 1971, 1974, 1976). The scales have repeatedly discriminated between various groups as predicted by the 10 theory. Further findings supporting the use of facet theory in pre- dicting correlational structures have also provided support for the construct validity of facet theory. Limitations of the facet-theory-developed series of ABS scales seems to be involved in the following areas: (a) response set; (b) social desirability; (c) homogeneous item content across the joint levels; (d) combination of facet elements; and (e) the effect of order of scale administration on correlation matrices. These shortcomings, especially (d) and (e) above, have been dealt with by Maierle (1969). By a theory Guttman means "A hypothesis of a correspondence between a definitional system for a universe of observations and an aspect of the empirical structure of those observations, together with a rationale for such an hypothesis" (Gratch, 1973, p. 35). This can be regarded as stating what Cronback has described as con- struct validity (Guttman, 1971). Guttman charges that it is the task of the social theorist to discover the structures underlying the totality of behavior: A task of the social theorist is to provide an abstract framework whereby to define the subuniverses; the more ade- quately it explicates the empirical correlations that ensue among the definitions, the better the framework. Compre- hension of the multi-variate system of the universe can lead to larger theories with relation to other universes and thus to more and more perfect multiple correlations for each variety of behavior separately. The improved predict- ability will not depend on mere empiricism, then, but will be guided by a systematic social theory (p. 318). Guttman (1959) further states that ". . . to improve the predict- ability would require enriching the facet design, or placing these 11 behaviors in a larger context" (p. 327). The research reported wherein further explicates the relationship between attitude objects- in-situation across three attitude levels. This was accomplished by measuring attitudes toward selected objects (Blacks, mentally retarded, and blind), holding constant the object-in-situation and the subject-object relationship. Maierle (1969) proposed the possibility that the two dimen- sions or structions, joint (that due to attitude level) and lateral (that due to content), interact in ways which were not accounted for in the present methods of simplex analysis and facet design. To date, no studies have examined the interactions of the joint and lateral struction; i.e., how the attitude levels and situations interact across various attitudinal objects. In summary, the purpose of the present study was to further examine the construct validity of the Guttman-Jordan facet theory approach to the measurement of attitude behaviors: (a) can the relationship be represented in a multi-dimensional model; (b) can the facetized theory discriminate between attitude objects; and (c) what is the joint-lateral interaction when the joint and lateral dimenSions are held constant across selected objects. Fieldman and Hass (1970) provide further justification for the need for the present research: ". . . psychological research paradigm should try to relate what different individuals do in a given situation, to what a given individual does under different conditions." In this project, the situations (life situations) were held constant and the different "treatment" conditions were 12 represented by three attitudinal objects. Furthermore, according to Jordan (1971), structioned variables should provide a set of clearly defined profiles fin~'tross-object, cross-situations, cross-national, and sub-cultural comparisons . . . ." And Krech, Cruchfield and Ballachey (1962) provide a final justification: "few attitudes exist in a state of isolation; most of them form clusters with other attitudes" (p. 145). In accordance with Guttman's proposal (1959, p. 327), an attempt was made to increase the facet design by incorporating more and wider facets of attitude and placing the behaviors in a broader context. Thus, an attempt is made to treat the theory as Jordan and Guttman (1976) suggest: "a theory for all variables simultaneously must account for variations over lateral and content facets, as well . . . . CHAPTER II FACET DESIGN, STRUCTURAL THEORY, AND ATTITUDE MEASUREMENT Definitions are, of course, arbitrary. Following Louis Carroll, one can make words mean what one wishes. Basi- cally, all that is formally required of a definition is that it be clear: that it enables reliable use of the concept concerned. A more formal, heuristic desideratum is that it actually influences theorists and researchers to progress in their work (Guttman, 1971, p. 329). Guttman (1955, 1959) developed facet theory as a tool or strategy in defining a research problem and in theory development. In social research there are usually two sets of variables: the population and a set of attributes or qualitative variables (Elizur, 1970). The attributes represent what Guttman refers to as the uni- verse of content of the investigation. Facet theory provides a means by which a systematic design of the universe of content is obtained, and therefore, facilitates the formalization of hypotheses regarding the relationship between the definitional system and the structure of the empirical observations (Elizur and Guttman, 1976). According to Guttman, the use of the structural approach to the development of psychological theory is becoming increasingly widespread. He defines a theory as: An hypothesis of correspondence between a definitional system for a universe of observations and an aspect of the empirichl structure of those observations, together with a rationale for such an hypothesis (Gratch, 1973, p. 35). 13 14 This definition emphasizes the necessity of defining the universe of observations to be researched, and also stipulates that the defini- tional system should be in a form that facilitates perception of correspondence with empirical data (Levi, 1976). Facet theory, as developed by Guttman and his associates, is a metatheory for the design of structural and other theories (Elizur, 1970). Furthermore, new innovations in non-metric measurement methodology also provide a means for quantifying the qualitative data of facet analysis and testing the structure of the data to indi- cate if it reveals the postulated statistical structure (Elizur, 1970). There are three basic constructs in facet theory: subjects or respondents (called the population and designated by P), the variables (attributes of the population), and categories (the sub- classes of the variables) (Kats, 1972). Each variable (facet) studied can be conceived of as a subuniverse of the total universe, where the total universe is all aspects of the universe (or the- oretical) problem. The collection of facets (variables) can be linked together via a mathematical statement of sets: ABC . . . . Each facet is conceived of as having "structs," members, or elements (See Appendix A, Glossary). The notation of facets is by capital letters and the elements of each facet are denoted by lower case letters. For example, facet A has elements a], a2, . . . am. Elements of the facets are combined to form profiles (structuples). For example, facets A and B may each have two elements and therefore there are four possible profiles: a1 b]; a2 b]; a1 b2; and a2 b2. The 15 universe of content is defined as the collection of all possible pro- files over the facets and their elements (Kats, 1972). Consequently, all variables included in the analysis form a universe of content and each variable can be defined as a profile of elements where each element belongs to one of the facets which defines the universe (Guttman, 1959; Elizur, 1970; Kats, 1972). In summary, each investigation concerns a set of variables (facets denoted by capital letters), a population (denoted by P), and a range of categories (responses denoted by R and expressing a common range). The combination of facets forms a profile, where each element is a "struct" and the elements together form a "struc- tuple." Structuples mean the same as "profile" (Elizur, 1970). , Thus, if a research problem has two facets, each with two elements, as in the example above, a structuple would be a1 b1 and each ele- ment would be a struct. To enable representation of a total design, Guttman devel- oped the technique of the "mapping sentence,“ which represents the relation in the following form: P AB +-R. In this mapping, the relations are mapped into a domain (population and variables) and a range (category of responses). The mapping provided above is a summary notation which says that for each respondent (an element, facet P), in a reaction to a question (a variable or structuple of facets) "implies" one answer in terms of categories (an element of y the range, facet R) (Kats, 1972). The arrow as used by Guttman does not imply a causal relation: only that "if what is specified in 16 the domain is true, the specification of the range is true" (Elizur, 1970; Kats, 1972). The complete mapping sentence is the equivalent of more formal expression as used in set theory notation; and the mapping sentence presents the complete research design (Elizur, 1970). Con- ceptually, the mapping of facets permits the inclusion of all facets that are theoretically possible, and all possible structuples. In order to facilitate communication and translation of the concept represented in the mapping sentence, Guttman proposed utilizing a standard grammatical sentence form by adding verbal connectives between facets (Guttman, 1965). In brief, Guttman's mapping sentence serves two purposes: (a) it provides a definition of the universe of obserations and (b) it provides the relation in a form that aids systematic percep- tion of the relationship (Levi, 1976). In effect, the mapping sentence is a basic technique in facet theory (Elizur and Guttman, 1976). By specifying in a mapping sentence the basic facets (or variables) which may, in part, influence, determine, or effect a response, a researcher is forced to thoroughly consider the aspects of his theory or research problem. Thus a strategy of "extension and intension" of theory is an important feature of facet theory: ". . . heuristic strategies are possible through mapping sentences, since they easily lend themselves to correction, deletion, exten? sion, and intension" (Levi, 1976). The advantage of facet theory lies in the capability of defining the components of a research 17 problem and in formalizing the process. In comparison to traditional factorial design, Guttman and Guttman (1975) state the following: . what may be regarded as a complete design for the pur- poses of analysis of variance (and other statistical analy- sis) turns out to be incomplete for at least two more basic purposes: (a) theory construction and (b) conducting the original observations which are to be subjected to the data analysis. The mapping sentence device is intended to make the experimental design more complete. In addition to the facets of the factorial design it brings out explicitly cer- tain other basic features required of the original observa- tion. The mapping sentence gives more specific instructions on how to make empirical observations, and in this sense provides the definitional framework for these observations. Given such a more detailed framework, it facilitates theory development (p. 3). Guttman's Post Hoc Facetization of an Attitude Research Problem In 1959, Guttman reanalyzed a design by Bastide and van den Berghe (1957), and abstracted, via facet theory, four subuniverses of attitudes which they had not explicitly designated. Bastide and van den Berghe had assessed interracial behavior in Brazil, and described the following four types of attitudes: stereotypes, norms, hypothetical interaction, and personal interaction. They had pre- sented the intercorrelations among the four types of attitudes. From a reanalysis of the data, Guttman developed a structural theory for intergroup beliefs and actions through the facet definition of the same universe of content. In his facetization, Guttman defined three facets; each, in turn, with two elements: 1. the behavior a1 beliefs a2 overt action 18 2. the referent to whom the behavior is ascribed b1 the subject's group b2 the subject h1mself 3. the type of behavior c1 comparative c2 interactive The Cartesian product of these three facets permits eight possible structuples with three structs each: a1 b1 c]; a1 b1 c2; . . . a2 b2 c2. An example of the a] b1 c1 reads: belief (al) of a sub- ject that his own group (b]) interacts (c2) with a specified atti- tude object. Similarly, the structuple a2 b2 c2 reads as follows: self or observed reports of a subject's overt action (a2) of him- self (b2) interacting (c2) with specified attitude object. There is an ordering of these facets in the design; Guttman refers to it as a progression from weak to strong forms of behavior vis-a-vis the attitude object. Ideally, within each facet the ele- ments can be ordered from weak to strong forms of behavior and the higher the subscript, the stronger the behavior. For example, overt action (a2) is stronger than belief (a1). This principle of order- ing has important implications which will be explored below. Thus, according to the facetization of Bastide and van den Berghe's data, eight subuniverses were possible: 1. a1 b1 c1 Stereotype: Belief (al) of a subject that his own group (does not excel) in comparison (c]) with Negroes 2. a1 b1 c2 Norm: Belief (al) of a subject that his own group (b]) ought (ought not) to interact (c2) with Negroes 19 l 2 c2 Hypothetical interaction: Belief (al) of a subject that he himself (b2) will (will not) interact (c2) with Negroes 4. a2 b2 c2 Personal interaction: Overt action (a2) of the subject himself (be) to (not to) interact (c2) with Negroes a1 b2 c1 Feel superior a2 b2 c1 Act superior 1 c1 Teaching CD \I 01 01 m N 0" b1 c2 Preaching As mentioned above, Guttman showed, through facet analysis, that Bastide and van den Berghe had intuitively arrived at four of the possible eight subuniverses and had not investigated the remain- ing four. Thus Guttman showed that, had the original investigators been guided by facet procedures in the design of their study and analysis of their data, they would have known in advance ". . . what to search for, how to test it, and what significance to subscribe to their results" (Elizur, 1970, p. 47). Structural Relations Specified by Facet Procedures Guttman's definition of a theory as quoted by Gratch (1973, p. 11) specifies a correspondence between empirical observations and a definitional system. Two related principles have been used in specifying this correspondence: the first is the proximity principle (formerly called the contiguity hypothesis) which states that: 20 subuniverses which are closer in their facet construction will also be closer statistically (Elizur, 1970, p. 58). According to this principle, the relationship between structuples decreases as the number of similar structs decreases. For example, eight possible structuples were specified by Guttman's facetization of Bastide's and van den Berghe's data. According to the proximity principle, the relation between the structuple a1 b1 c1 and a1 b1 c2 would be higher than the relationship between a1 b1 c1 and a2 b2 c2. However, the real relationship depends on the theoretical weight given by the "facets, elements, or their combinations" (Kats, 1972, p. 41). The second principle is that of structural order. In this principle, ordered proximities (structures) is important. Order is presumed to exist between elements of facets (structs) and between different structuples. As the structuples combine with increasingly stronger elements from the facets, the total structuple can become stronger. For example, stereotypic is weaker than personal inter- action, and personal interaction is the strongest profile which . Bastide and van den Berghe defined. The concept of order or lawful- ness leads to prediction of empirical structures from ". . . consid- eration of order within elements of the facets concerned" (Elizur and Guttman, 1976, p. 2); lawfulness here refers to geometric proper- ties of obtained correlation matrices. Since the mapping sentence enables one to project the empirical relationship between structuples, Guttman and his asso- ciates have discovered several forms of structural relationships. 21 The discovery of structural order or lawfulness is one of the chal- lenges to the social psychologist: Recognizing that differential relations exist within and between variations of behavior, the challenge to the social psychologist is to reveal what structural system, if any, underlies all of these relations (Guttman, 1959, p. 318). Since Guttman and associates have found that correlation structures remain relatively unchanged over time and circumstances (Guttman, 1964; Guttman and Levi, 1970), whereas means or averages are sub- ject to considerable variation, it seems only reasonable to expend considerable effort to discovery of structural lawfulness. The "simplest" form of structural order is the simplex (Guttman, 1954). The simplex is a statistical structure of inter- correlations which reveals a "simple order of complexity" (Guttman, 1954, p. 260). The simplex is determined by the fact that the "highest correlations lie along the main diagonal where the features are closer together in their a priori order and taper off toward the upper right and lower left corners of the matrix, where there is the greatest difference in the a priori order" (Guttman and Guttman, 1965, p. 220). Table 1 portrays a simplex structure with four variables. In his search for a single-common-factor which shows order among variables, Guttman states the following about simplex structure: Suppose we are given n tests t1, t2 . . . t which differ only on a single complexity factor . . . . Test t] is the least complex. Test t2 is the next; it requires every- thing t1 does and more. Similarly, t3 is more complex than t2, requiring everything t2 does and more . . . . In this 22 TABLE l.--Simplex Structure of Correlation Matrix.1 Variable l 2 3 4 l 1.0 High Lower Lowest 2 High 1.0 High Lower 3 Lower High 1.0 High 4 Lowest Lower High 1.0 l Adapted from Elizur (1972, p. 59). case, t3 is also clearly more complex than t]. In general, test to + l is more complex than t-, and hence requires what all preceding tests require p1us something more. Let G denote the total complexity factor, of which all the tests are composed in various degree. Thus, G is like an additional test beyond the most complex given test tn . . . (1954, p. 269). Given the order of complexity, it seems only reasonable to expect that the correlation between t1 and t2 would be higher than between t1 and tn, given the fact that t1 and t2 differ only in order of complexity. Thus Guttman (1954) is able to specify the perfect simplex. In reality, the perfect simplex rarely exists, but the principle of order among the variables is finding increasing support (Brodwin, 1973; Elizur, 1970; Gottlieb, 1973; Guttman, 1954; Jordan, 1971; Kats, 1972). This simplex structure has been shown to be a factor pattern (Guttman, 1954). In a geometrical sense, a simplex may be thought of as a collection of points along a straight line (Elizur, 1970); in content, the simplex implies differences of degree (Kats, 1972). 23 The circumplex structure involves the law of proximity as does the simplex (those structuples closer to each other in their facet design will correlate higher), but there is a circular order to the correlations: i.e., "a circular order of complexity" (Gutt- man, 1954, p. 260). In the circumplex the strongest correlations occur along the main diagonal and decrease in size as they move away. However, toward the corners of the matrix the correlations increase again, which makes the total picture one of circular order. It is possible to see a circular order among the variables by direct inspection of the coefficients of monotonicity (the correla- tion coefficient). Table 2 provides a hypothetically perfect cir- cumplex structure of correlations. TABLE 2.—-Intercorrelations for an Equally Spaced, Uniform, Perfect, Additive Circumplex. f Test t1 t2 t3 t4 t5 t6 t] 1.00 .75 .50 .25 .50 .75 t2 .75 1.00 .75 .50 .25 .50 t3 .50 .75 1.00 .75 .50 .25 t4 .25 .50 .75 1.00 .75 .50 t5 .50 .25 .50 .75 1.00 .75 t6 .75 .50 .25 .50 .75 1.00 Total 3.75 3.75 3.75 3.75 3.75 3.75 1 Adapted from Guttman (1954, p. 329).. 24 In a perfect circumplex, the column totals will be equal. A quasi-circumplex is defined as a perfect circumplex plus deviations (Guttman, 1954); and a quasi-circumplex is in reality usually obtained. The radex involves both the simplex and circumplex. Given two facets in a design where facet A has three levels and facet B has three levels (i.e., there are nine possible structuples), facet A may be an unordered facet and facet B an ordered facet. If these constraints are true, a radex would be the expected outcome. Fig- ure 1 provides a pictorial representation of the hypothetical radex, where facet A plays a polarizing role and facet B a modulating role. The polarizing effects of elements of facetA is to separate the al. Figure l.--A Diagram of a Hypothetical Radex. 25 space into regions, each of which emanate from the origin and radi- ate outward, each in its own direction. The modulating effect of the elements of facet B is to modulate the distance from the origin. If one were to hold constant one of the elements of facet B, then a circumplex would result. Likewise, if one were to hold constant one of the elements of facet A a simplex would result. Thus a radex is a form of lawfulness which Guttman called "radial expansion of com- plexity" (1954, p. 260); in the example provided above, facet 8 elements formed concentric circles and facet A elements formed the segments. A cylindrex or a three-dimensional representation of correla- tions is defined as a two-dimensional radex and an axis orthogonal to it. A radex is a circular arrangement in a plane, and the axis perpendicular to it defines a cylindrical configuration. An axial element or facet would also involve elements which would be ordered, and the orders would be represented along the axis of the cylinder (Levi and Guttman, 1975). Therefore, three facets are required to form a cylindrex. Two play roles in the radex; one would polarize and the second would modulate, and the third facet specifies orders along the axis. Figure 2 provides geometric representation of a cylindrex with three facets, each having three elements. As in Figure l, facet A is a polarizer and facet B a modulator and facet C acts as an axial factor. These structural hypotheses have received wide support: in mental abilities (Guttman, 1964), worry (Levi and Guttman, 1975), and attitudinal measurement (Jordan, 1971; Jordan and Guttman, 1976). 27 Other structural relations and lawfulness are currently being obtained and evaluated at the Israeli Institute of Social Research. Indeed, Guttman's 1954 speculation about other possibilities of fur- ther lawfulness, given the notion of order among the variables, implied a facet design. Also, once we focus on the notion of order amongst vari- ables, alternative theories of order are possible . . . even in a plane . . . a part from those of the symplex, circumplex, and radex. One can imagine ex-strings of the elementary components, with loops in them, etc. (Guttman, 1954, p. 240). Nonmetric Analysis: Examination of Structural Order and Lawfulness Guttman's definition of a theory (in Gratch, 1973, p. 35, and quoted on page 13) states that a theory is a correspondence between a definitional system (facet theory in this research) and the empirical structure of observations, together with a rationale for the hypothesis. His definition emphasizes that the structure of the relations will be specified by the facet design and as shown above by the ordering principles. Guttman (1968) and Lingoes (1973) have developed a series of nonmetric techniques which have proven to be useful in portraying the structure or lawfulness involved in cor- relations or other "distance" functions. Most of the earlier methods used to study relationships between many variables are subject to the constraints (assumptions) of least squares analysis. The Smallest Space Analysis (SSA) method ' departs from strictly metric assumptions and replaces these by ordi- nal assumptions. SSA-l is the first of a series of methods which are rr—-—.-.-i 28 based on ordinal distance models (Euclidean geometric relations) for the analysis of data matrices; these apply transformations of coef- ficients of monotonicity. Thus, SSA-1 is specifically designed for spatial representation of symmetric matrices of similarity or dis- similarity coefficients, such as correlation coefficients (Guttman, 1966, 1967, 1968; Lingoes, 1965, 1966, 1968, 1973; Lingoes and Ros- kam, 1971). SSA-l calculates coordinates for points representing vari- ables such that the distance points reproduce the rank order of the association values (between variables) according to a criterion of fit--the monotonicity criterion; and reproduce the smallest possible dimensionality in a Euclidean space. In its simplest sense, SSA-1 portrays physical and spacial distance between variables as repre- sented by the correlation coefficients. For example, if the corre- lation between variables X and Y is +1.0, then they would occupy the same space. If the relationship was -l.O, then they would be far apart in the space. Adding the correlation of X and Y with a third variable, Z, to the space would result in SSA representing the dis- tance between all three variables. If the correlation between X and Y was -l.O; between X and Z, 0.0; and between Y and Z, 0.0, then a one-dimensional space of a straight line would represent the rela- tions. The result is a configuration of points (variables) based on the sizes of the correlations or other distance functions between all variables. Reproduction of the values of the correlations is not the sole aim; only ordinal restrictions are imposed upon the solution, which is why it is called a nonmetric approach (Kats, 29 1972). The configuration of plotted points is essentially the objective of the analysis; this contrasts with factor analysis in which the coordinates of each point are interpreted as factor load- ings after acceptable rotations of factors have been determined. Thus, smallest space analysis attempts to use as few coor- dinates as possible, a minimal number of dimensions, to acquire an adequate representation of the rank order of relations and the con- figuration of points: the smallest space. The program searches in an iterative process for the most adequate configuration of points in that number of dimensions of smallest space. Since fewer dimensions are needed to reproduce order infor- mation than metric information, SSA-l results in a simpler and more direct data representation, and therefore is viewed as a more parsi- monious method (Guttman, 1966; Lingoes, 1966). It is also more par- simonious than factor analysis (Schlesinger and Guttman, 1969). While factor analysis and smallest space analysis will produce the same basic structure of data (Kats, 1972), SSA usually renders fewer dimensions than factor analysis, i.e., a smaller space than factor analysis of the same data (Elizur, 1970). In one example, a smallest two-dimensional space was equiva- lent to a six-dimensional factor space (Guttman, 1966). Schlesinger and Guttman (1969) reanalyzed existing data in which factor analysis (by the orthogonal method) had obtained a six-factor space. Through SSA-l they found that the data could be represented in a two-dimensional space and preserve the basic structure obtained in g (1‘ 91m A" J a I 3O factor analysis. From this comparison with factor analysis, Schlesinger and Guttman (1969) concluded the following: 1. Smallest Space Analysis makes it possible to arrive at a smaller space than does factor analysis. In the empirical example of this paper, a two-space has been shown to portray adequately data for which six factors had been extracted by factor analysis. 2. The configuration of points revealed by Smallest Space Analysis corresponds essentially to that yielded by factor analysis. Indeed, the factors extracted by factor analysis can be represented by points in the smallest space. 3. The notion of coordinates is not essential. It is suggested that an analysis of test content in terms of definitional facets may lead to more fundamental insights into laws of formation of the structure of correlation matrices. In SSA-1, the concept of dimension has nothing to do with the content of the data; it represents the smallest space in which the configurations can be shown. The facet design, taken together with the ordering principles and structural hypothesis, attaches meaning to the obtained structure. The tendency to look for meaning in the dimensions and coordinates of the dimensions is virtually meaningless without taking into account the facet design and content of the data. In contrast to factor analysis the meaning is attached a priori, via the facet design. The coefficient of alienation shows the degree of fit of the solution and measures the deviation between input coefficients and the reproduced distances. This coefficient ranges from O to l in such a way that the better the fit between the data matrix and the configuration, the closer to zero it becomes. For a two-dimensional space, a fit of .15 is thought acceptable, but for several reasons is not an absolute criterion. A more important guideline is the 31 interpretation of the configuration in terms of facet design and content; at times the rule of a technically perfect fit has been violated in favor of content interpretation. SSA-l analysis reflects a quest for ordered structures; its interpretation stresses the configuration of data and rank order among relations rather than their absolute size. These ideas were 1_1 first elaborated upon by Guttman in radex theory, with facet design 5 as the theoretical framework for predicting and interpreting such ‘1 data structures (Guttman, 1954, 1959, 1965, 1966). Generally, the interpretation is done graphically. Facet Theory and Attitude Measurement While Guttman proposes that "all of human behavior towards social objects can be divided into subuniverses . . .“ (Guttman, 1959), he is more concerned with specific patterns of behavior than the possible underlying characteristics of individuals (Maierle, 1969). As shown above, facet theory is a method for the design of structural and other theories and new innovative nonmetric method- ology provides a means for facet analysis and for testing of struc- tural hypotheses generated by the facet design (Elizur, 1970). This is truly the sense of a theory as Guttman defines a theory. Guttman (1959) distinguished three "facets" involved in a particular attitude response: facet A, the subject's behavior (a1 belief vs. a2 overt action); facet B, the referent (b1 the sub- ject group vs. b2 the subject himself); and facet C, referent behavior (c1 comparative vs. c2 interactive). He further postulated 32 an ordering principle from weak to strong forms of behavior: i.e., the elements of the facets are ordered and as the structs become stronger, the strength of the structuple becomes stronger. Thus, all attitude items can have none, one, two or three strong structs; a total of four possible combinations from weak to strong structs. Guttman's theory showed a logical reason for only four permutations. If the elements of the facets are properly ordered within each facet and the facets are correctly ordered with respect to each other, then analysis of attitude items by n-dichotomous facets will produce n + 1 types of attitude items. These types are called "levels" where each "level" has one more strong element than the "level" preceding it and one less strong element than the "level" immediately following it. In Guttman's reanalysis of Bastide and van den Berghe's (1957) attitude research, Guttman arrived at eight subuniverses, at four levels. Thus, Guttman's (1959) facet analysis of Bastide and van den Berghe's data allowed for three facets and hence four levels of atti- tude. Guttman (1959) also suggested that to increase the predict- ability of his theoretical model, it would be beneficial to (a) enrich the facet design and (b) place these behaviors (levels) in a broader context. In the latter 19605 and early l970s, further application of facet theory to attitude measurement was undertaken by Jordan and others at Michigan State University. At first, Jordan. utilized a facet design to construct a universe of attitude items toward the mentally retarded (Jordan, 1968). Jordan (1968) projected 33 that there were other pertinent facets, and accepting those identi- fied by Guttman, he expanded the facet analysis of attitudes to include five facets and therefore six levels. Table 3 provides Guttman's original facets and Jordan's adaptation. Specifically, Jordan added facets which defined two more levels at the lower end of Guttman's original levels (Brodwin, 1973). According to Guttman, an item (a structuple from a complete mapping sentence) belongs to the universe of attitude items if the following constraints are true: an item belongs to the universe of attitude items if and only if: its domain asks about behavior in cognitive . I I affective modality toward an object, instrumental to and its range is ordered from very negative 3very positive: toward that object (in Gratch, 1973). Guttman's original attitude levels (stereotype, norm, hypothetical interaction, and personal interaction) were primarily concerned with the cognitive and affective modalities (McGuire, 1969). It is at this point that Jordan visualized the need to expand Guttman's sys- tem to include conative modalities; his six-level facetized design gave greater emphasis to the affective and conative elements of "attitude behaviors" than does Guttman's original four-level design, thus the term "attitude-behavior." Jordan's additional levels actually emphasized real, observable, experienced, and/or reported behavior. These levels evaluate the subjects' actual feelings and actions, instead of perceived thoughts, beliefs, and opinions. 34 upmmsm; :owuum 1 1 . o>wuumcmucm we puwnnzm No acm>o Na 1 l . azocm I I . m>wamcmaeou Pu m.uumwa:m pu mwwpmn Fa . u I co_>m;wn gnocmcmpcw sow>mgma 1 - . m.ucmcm$mm pcwcwemm m.uumnn:m . u . casuuau Acow>mgmn A»E\mcwev ucm>dq. AHV Pacowumemao mm comuomcmucw No upmm Nu mucmwcmnxm Na upmm mm _muwpm;uoazs pm comwcmasou Fu mcmcuo Fu mmPFma Fa msmcuo pm Lop>msmn som>mcmn m.couum asogmcmpcw Lom>mzmn we :_meoo m_couu< cogu< ucmcmwmm ucmcmemm cause: u a u m < 5.58 cowumuamu< cmucoa cm mumumm .memuH mvzgmuu< com mcmwmmo pound cmucoa can cmsupzo wo comwcmnsou-i.m m4m
mp maowgu> No mn_;m=owum_oc muouwvcw mmmozucmcoa cm muse: mumccmu_<
.mucmsmumum acmumpmcou aan «emanates we “can use mmmogucmcma cw mucosa
.Fm>op cw mucoem_m acocum ma Logan: u .02
.m anah .eua
.mm< as» cw new: mcovum:_nsouse
u
a
appocopumcomo Acow>ugon NmNuNUNnNm
cowuuo .mcomcma «a uco>ov mcowpuocmuew Na mucmwcmaxw.m m NF m w e o m o
sppauwsmguoaxm _5Nu~u~n~a
mc_—mmm _mcomcmg as Ammcwpomwv mcowuuocmunw Aalmucmwcmaxw.m v o— c p E o w m
appoumumguoaan ; u o m o
cowuum aaogm pmauu< mcompgaqsqw .mcmcum mucmwcmaxm meccam ——
.:o_uum m FwNuNUFnNm e
—uu_um:uoaxc _m:omcma «a appouvumguoqam mcopuumcmuaw am m>ompom.m m g P E n P
Ammcw_mmm asogm .1 .1 x_puuwuwmuoaxm.1
pmauuav xuwucmu_ aaogo meowuumcmpcw .mcmguo mucowcmaxw mgosuo m z w o o o
Amcopumuuma .1 .mppmup.1 p a a o
1xm qaocmv mzo— noswmpuoca uumguonxm mcopuumcmucw am m>mwpma mcmzao m g .
Amauuum pccomcmq vmcmwmmm appmuwumguogam mcommcmaeom ha m>mmpum.m N N g u e a w m
appmcomconv unmucouimpmm p N p F N
Amm:_o> um>wmucmqv .1 .1 appmyw.1 o v u a a
:owuozpm>m Pecos Focomcma «« lawsuoqu mcowuuogmucw .mcmzuo m>awpma H c z w o a P
.1 N..flwium;s.1 ; u s a o
mzuuum lonam mcompgunsou ha m>mmpmn mcmcuo m
—~comcma umcmpmmmiazocu .mppmumummuoaxm.1 pmNupupapa
.Eo: Zufiuom 2.. 233235 .9858 26.53 22.8 F N Airlolalo N
magnum 1. APquwumsw.1 .
naocm umcmwmmm xppmcomcoa -oaxm mcomwcoasou .mcmguo m>mmpwa H m g u o a w
Amadeus _ F F _
aaogm nmcuwmmm1a=ocmv 1. .mppouwummuoaxm.1 Fm u u n a F
maxpoocmpm .aumvuom «a mcomwconsou .mcoguo o>m_pma mcmsuo o F g u o a o
umsaz a>_uaitum~a oucaeauaum _~=o_sicccma 5.52 _mpmmuup owwwwua Pm>mn
.mcopuucvnEou m>pmzh no; mandamumum
_a=o_uwe_tmo ace .mm__COLa “mama m_a>mn " mcossa~aianto> musspsu< to smumsm p5>wu-x_m umumu-a>ca--.o mnm 84 mmocu< map—mom =o_um=—o>m page: UNQAuomcmpm
.umsciucou--.¢_ mnm m— No.0“ .0. .uumhno any "x—m>.uumommc .mcwoo c. .ucmmmcomc mcouum— xpm as» w.no.co> zoom c.n
.uoqooco coma o>og m.oE.umoo
- oN so oN N. .o .. No No- mm .N on oo- No- oo o. no. oo o. oN .o .o .o so- m. so- No- omNNoN .N
- mm No m. .o- o. .N o. .. mm on s.- so no mo- No No No mm os No- No so. No oo oo Nmaaoo oN
- s. N. oo o. .. o. m. mN sm so- No mo- mo- oo ..- NN .m on s.- No. s.- so No- NN- NNENoN mN
- ms ms No N. oo o. No- so- NN .o- .o- .. o. N. .. mo- .o- sN so No NN oo- s.- omNzoo sN
- No No Ns NN .. mo mo o. NN o. o. No NN s. N. mo m. .m .N o. N. No NmNzoo NN
- mo- .m on No- m.- .o- No mo o. o. o. s. oo mo- No mN .N oN o. .o oo- Naomoo NN
-- NN o. so so No- NN so .o- NN No- No- No o. N. so- No. so- N. so- so- om.moo .N
-- oo oo oo o.- No- o. mo N. on o. No No o.- so o. s. No oN o. mm.moN oN
- .o- No mo- No- N. o. s.- s. o. No .o- .o- mo- oo oo oo No mo- NN.m.o o.
- mN om No so- oo mN .o- oo Nm N. so mo- s.- ..- .. .o No- omNNNz o.
- ss o.- No s. No No oo .N ms om mo- No- s.- No- N. m. NmNNNz N.
- o.- mo- m. No N. o. os om Ns .N- N.- ..- oo- so. so- NNENNz o.
-- om oN m. No- .o N.- o.- .o .N s. o. s. oo- mo- omNzNz m.
- ss N. oN o. No- No so om om oN No- m. .. mmNzNz s.
- NN Nm Nm mo- oN .N mN NN os mo NN mN NNNZNz N.
-- oN m. N. no mo- so No- oo- o. m. N. om.maz N.
-- os mo- .o- oo s. NN mN No om NN mm.mNz ..
- s. so- mo NN s. oN . so NN oN NN.mNz o.
- Nm N. mo- .o oo- s. oo- so- omaaso o
- oo mo- o. s. o. o. No umaaxo o
- so No oN oo- No mo NNNNEN N
- om .s mo oN oN omngo o
- o. so- Ns on mmozxo m
- mo- ms om NNNsz s
- N.- N.- om.mxo N
-- .. om.mgo N
- Na.mxo .
NN oN mN sN NN NN .N oN m. o. N. o. m. s. N. N. .. o. o o N o m s m N . oo..s.cs>
.muumwno ouauwuu< new pm>m4 umcosuuocum u:_oo an mucoumnom mucopmwo —m_uom Lo» omcowuo.mLLoU11.m. m4m