1' ||||HUl|l|l)IIHIIHIHIHHllllllillfllilflflljlflflllill ‘ 3 3 1293 1043 , -. SW70 6; 23L '3" 9* 0321995 A MULTIDIMENSIONAL SCALING ANALYSIS OF GROUP MEMBERS' INTERFERSONAL PERCEPTIONS By Stephen William Grimm A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Counseling, Guidance, and Personnel Services College of Education 1977 ABSTRACT A MULTIDIMENSIONAL SCALING ANALYSIS OF GROUP MEMBERS' INTERPERSONAL PERCEPTIONS By Stephen William Grimm Group members' interpersonal perceptions are the topic of the study. Three theoretical areas merge in the field of interpersonal perception. These include theories of perception, theories of inter- personal behavior, and small group theory. An integrative focus on these three theoretical areas was used to approach the study of inter- personal perception in groups. Multidimensional scaling techniques were used in the study. Previous research using multidimensional scaling to analyze interper- sonal perceptions in groups was reviewed. Complex judgments were found to be structured parsimoniously along a relatively small number of dimensions. The identification of these dimensions was hindered by the lack of a comprehensive external criteria that was needed to label the dimensions given by a multidimensional scaling analysis. The aims of the study were, first to describe the best fitting dimensional representation of group members' perceptions, second to label these dimensions, and third to compare group leader and partici- pant subject weights on each dimension. Data were collected from five Stephen William Grimm interpersonal process groups that were affiliated with psychology and education courses at Michigan State University during Winter term 1977. Four groups had eight members, and the remaining group had nine members. Measures were administered during the eighth week of the groups' nine- week duration. Two types of measures were taken on seven-point Likert scales. Subjects completed similarity ratings of all possible pairs of group members in their own group, and also rated each group member, including themselves, on sixteen adjective pairs of a Group Semantic Differential (GSD). The sixteen adjective pairs were chosen to describe four factors: dominance, affiliation, activity, and goal orientation. A factor analysis and elementary linkage analyses were performed post. Egg_on the GSD adjective pairs to check the construct validity of the proposed factor clusters. The factor analysis yielded a four-factor solution and a three-factor solution that both modified the proposed factor clusters. The linkage analyses showed that each group contained a distinct pattern of adjective-pair clusters. The GSD was designed to be an external criteria for labeling dimensions derived from the simi- larity ratings. The proposed clusters of adjective pairs, the clusters defined by the two factor analysis solutions, and the clusters described by the linkage analyses were used for further analysis. An individual differences multidimensional scaling analysis (INDSCAL) was used to analyze the similarity ratings made by each subject. Three dimensions accounted for between 63 and 84 percent of the variance in similarity ratings across the five groups. The three-dimensional Stephen William Grimm solution was chosen as best representing subjects' similarity percep- tions. Spearman rank correlations were computed between the INDSCAL dimensions and the GSD ratings in each group. The results indicated that the labels of the dimensions varied from group to group. The adjective—pair clusters most highly correlated with the INDSCAL dimen- sions were used to identify the dimensions. The amount of total simi- larity variance recovered by the GSD clusters from the INDSCAL dimensions was approximately 50 percent in each group studied. Four of the INDSCAL dimensions across the five groups were not well identified. The subject weights of leaders and participants on the INDSCAL dimensions were com— pared. Generally, leaders weighted activity heavier than did partici- pants, and participants weighted dominance heavier than did leaders. The variations found in the attribute selection for similarity judgments in the groups are discussed and highlighted the applicability of a domain-specific dimensional model. Dedicated to James W. and Eileen P. Grimm for their mutual interests in people and performance. ii ACKNOWLEDGEMENTS Special thanks are appropriate with the completion of this project. First of all I extend my appreciation to the members of my guidance committee: to my chairman Dr. William C. Hinds, for his organizational skills, clear evaluations of my writing, and overall commitment and caring about my personal and professional development; to Dr. Sam Plyler, for providing a model of actualized values that encouraged and supported me through my graduate education; to Dr. John R. Hurley, for initially stimulating my interest in group research and interpersonal behavior that enabled me to produce this final project; and finally, to Dr. Patricia Busk for her aid on methodological aspects of the study and for her efforts spent critically reviewing this document that modeled a sense of personal responsibility and integrity. I also extend my gratitude to Dr. Bob Wilson and the Education 200 staff at Michigan State University. Most particularly I thank the group leaders and members for their willing participation in the project. To my family, for its support and trust, I am ever grateful and proud to be a part of it. To Justin and Bryan, their presence gives me simple meaning and timeless satisfaction. Finally, to Jeanne extends my deepest appreciation and affection for not allowing the sacrifice of much time and energy take away from our bonds with one another. iii TABLE OF CONTENTS CHAPTER Page I. THE PROBLEM00000000000 ....... .00... ..... 000.000.000.00... [—4 IntrOduction 0 . 0 0 0 0 0 0 0 . . 0 0 0 0 0 0 0 0 0 0 . 0 0 0 0 0 . 0 0 0 . 0 0 0 0 0 0 0 0 0 0 Purpose. 0 . 0 0 0 0 0 . 0 . 0 . . . 0 0 0 . . . 0 0 . . . 0 0 0 0 0 0 0 0 0 0 0 0 . . 0 . 0 . 0 0 0 ResearCh Questions . 0 0 0 0 0 . 0 0 0 0 0 0 . . 0 . 0 . 0 . . 0 . 0 . 0 0 0 0 . 0 . 0 0 0 Theory.0.000.000.000.000000000000.0.....0....0.....0.. overv1ew000000000000.00000000000000000.0.0.0...0000000 1 «L‘bbUH I I . REVIEW OF THE LITERATURE 0 . 0 . . 0 0 0 . ..... . . 0 0 0 0 . 0 0 0 . 0 0 0 . . . 0 0 15 sumry000....0.....0..00..00.0. ...... 0.00.00.00.00... 25 III 0 DESIGN OF THE STUDY 0 0 . 0 0 . . 0 0 0 . 0 0 0 0 . . . . 0 0 0 0 0 . . 0 0 0 0 0 0 . . 0 0 0 . 29 Sample................................................ 29 Psychology Groups.................................. 29 Education Groups................................... 31 Measures.............................................. 34 Similarity Ratings................................. 35 Group Semantic Differential (GSD).................. 36 Research Questions.................................... 53 Analysis of the Data.................................. 54 Summary............................................... 57 IV. ANALYSIS OF RESULT8000.00.00.00.000000000000.00.000.00... 59 Dimensional Representation............................ 59 Dimension Labels...................................... 62 Subject Weights....................................... 76 Summary............................................... 79 V. SUMMARY AND CONCLUSIONS.................................. 81 Summary............................................... 81 Conclusions........................................... 84 Limitation of the Study............................... 87 Discussion of Findings................................ 89 Implications for Future Research...................... 95 BIBLImRAPHY0000000000.....00...0.0.0.0000000000000.000.00.00... 99 APPEmeA.0.00.0...0.00.00.00.000000000....00.00000000000000... 103 iv TABLE 2.1 LIST OF TABLES AND FIGURE Dimensions of Interpersonal Perception Identified Through Multidimensional scaling...0.0.0.000...0000.00.00.00000000 3.1 Correlations of GSD Adjective Pairs Across All Groups (n:41)....0...0000.000.000.00...00.00.000.0000000000000. 3.2 Varimax Rotated Factor Matrices and Proportion of Variance 3.3 Expected Adjective Pair Clusters and Adjective Pair Clust- ers from Factor Analys180000000000000000000.000.000.000... 3.4 Elementary Linkage Analysis Adjective Pair Clusters for EaCh Group00.0000.0.0.0.0...00000000000.000000000000000000 4.1 Average Subject and Mean Square Correlation Coefficients Between Computed INDSCAL Scores and Original Similarity Data000000000000000.00...0.000.000.0000000000000000.00.... 4.2 Spearman Rank Correlations Between the INDSCAL Dimensions in Each Group and the Sixteen GSD Subscales............... 4.3 Spearman Rank Correlations Between the 3 INDSCAL Dimen- sions in Group A and the GSD Adjective Pair Clusters...... 4.4 Spearman Rank Correlations Between the 3 INDSCAL Dimen- sions in Group B and the GSD Adjective Pair C1usters...... 4.5 Spearman Rank Correlations Between the 3 INDSCAL Dimen- sions in Group C and the GSD Adjective Pair Clusters...... 4.6 Spearman Rank Correlations Between the 3 INDSCAL Dimen- sions in Group D and the GSD Adjective Pair Clusters...... 4.7 Spearman Rank Correlations Between the 3 INDSCAL Dimen- sions in Group E and the GSD Adjective Pair C1usters...... 4.8 Average and Range of Leader and Participant Subject Weights of INDSCAL Dimensions in Each Group............... FIGURE 3.1 Diagrams of adjective pair typal Structures for all groups Page 27 41 42 43 45 60 63 65 67 69 71 73 78 47 CHAPTER I THE PROBLEM Introduction Interpersonal perception involves how people experience their human environment. The stimuli in interpersonal perception are people and their interpersonal behavior. The internal representations employed by subjects in interpersonal perception have long been of concern in social psychology (Tagiuri, 1958). The importance of individual differ- ences in perceptual processes was initially professed by the Gestalt school of psychology. The Gestalt emphasis on perception broke ground for later interest in person perception. Hastarf, Richardson, and Dornbusch (1958) suggest that the relevance of the field of inter- personal perception lies in its relation with interpersonal behavior. The study of the categories of interpersonal perception is primarily important as a prerequisite for investigating the vital problem of the extent to which knowledge of perceptual categories will provide us with a tool for understanding interpersonal behavior in general (p. 59). Cronbach (1958) proposes another reason for the significance of research in interpersonal perception. Until we are in a position to standardize judgments of particular‘others in manner proposed here, (on stimulus properties) ... judgments will be confounded with constant differences between judges as perceivers (p. 366). Cronbach suggests that ignorance of a judges perceptual map will be a roadblock to the analytic treatment of interpersonal perception and behavior. A mutually shared perception of interpersonal interaction and the perceptual representation of the field forms a complex domain of many elements. Cronbach (1958) states that ... theories of perceptual response should take into account the traits being perceived, the constant tendencies in this perceiver with respect to those traits, and finally the effect of the particular other as a social stimulus to the perceiver (p. 375). Bieri, Atkins 25 a1. (1975), outline four elements: (8) input, or stimulus variables, (b) output, or response variables, (c) characteris- tics of the perceiver, and (d) situational variables. A coherent theory and suitable methodology in social perception must attend to attributes of the perceiver, attributes of the person being perceived, and attri- butes of the situation that the interaction is embedded within. Research methods in the field of interpersonal perception must take into account a broad range of individual styles of perception. Traditional sociometric methods do not identify attributes underlying a subject's perceptions (Jones and Young, 1972). Many investigators (Hartley, 1968; Bales, 1950, 1970) have resorted to selecting attributes §_priori. Jackson (1969) states that in the domain of personality theory and interpersonal processes little is known about attributes underlying interpersonal perception due to scarce criteria for goodness of fit. Consequently, Jackson (1969) warns against specifying attri- butes to subjects. The arbitrary selection of variables results in the loss of an important source of variability. Differences between indi- viduals in regard to the attributes they choose to use in perception are neglected (Bryson, 1974). Early investigators (Taguiri, 1958; MacLeod, 1958) in the field of person perception promoted a phenomenological mode of study. MacLeod (1958) defines phenomenology as "the systematic attempt to ob- serve and describe in all its essential characteristics the world of phenomena as it is presented" (p. 34). This approach centers on the phenonemenal data of self, other, and the social context. An assessment of individual differences is possible. Attributes most relevant to sub- jects perceptions may be described. In the present study, attributes most relevant to subjects in their perceptions of other members of a small group are analyzed. Purpose The purpose of this study is to describe the interpersonal per- ceptions of subjects in an interpersonal process-oriented group. The identity and importance of attributes underlying subjects' perceptions are analyzed in three steps. (a) Subjects' similarity/dissimilarity judgments of all group members are analyzed through a multidimensional scaling analysis. (b) Semantic differential scales are used in an attempt to externally define the attributes derived from the scaling analysis. (c) the importance of the attributes in group members' per— ceptions is examdned in relation to the member's role as leader or participant in the group. The study analyzes interpersonal perceptions as they relate to the interpersonal relationships that comprise a group's social structure. Research Questions The aim of the study is to identify the attributes most relevant to group members' interpersonal perceptions. The objectives are clari- fied in the following research questions. Question 1. How many attributes are used by subjects in making similarity judgments? (Question 2. What are the identities of the attributes derived from the subjects' similarity ratings? Question 3. Do group participants and group leaders differ in regard to the importance of particular attributes in making similarity judgments? EM Studies in interpersonal perception investigate characteristics of the perceiver, personality attributes of the person being perceived, and the structure and dynamics of groups. The nature of interpersonal perception dictates a review of these three areas of theory. Fiedler (1958) points out that research in interpersonal perception provides "an integrative focus for three major fields in psychology: perception, personality theory, and small group behavior" (p. 243). Theory in each of these areas will be discussed, beginning with attributes of the perceptual process. The study of interpersonal perception in American social psychology may be traced to the Gestalt school of psychology. The basic assumption, that a social relation depends upon interpersonal perception is derived from the Gestalt school. Snygg and Combs (Kelly, 1955) have formalized this assumption into the postulate, "all behavior is determined by and pertinent to the phenomenal field of the behaving organism" (p. 40). Jones and Young (1972) have defined interpersonal perception as "both instrumental to social interaction and conditioned by it" (p. 108). The phenomenal field of interpersonal perceptual processes is characterized by inter-dependence of perception and action. Kelly's (1955) theory of personal constructs is derived solely on "creative the basis of man's perceptual processes. He emphasizes the capacity of the living thing to represent the environment, not merely respond to it" (p. 8). A "construct", defined by Kelly as "a way in which some things are construed as being alike and yet different from others" (p. 105), is the means through which a person represents their environment. The fundamental postulate of Kelly's (1955) theory is that "a person's processes are psychologically channelized by the ways in which he anticipates events" (p. 46). Consequently, a construct's con- tinuing validation and maintenance depends upon its predictive effective- ness. This process is referred to by Kelly as "constructive alternativ- ism" and assumes "that all of our present interpretations of the uni- verse are subject to revision or replacement" (p. 15). Kelly views man as a scientist, contemplating and understanding in his own way his stream of experience. The effects of Kelly‘s (1955) postulates in the field of inter- personal relations are specified in the "commonality" and "sociality" corollaries. The commonality corollary states that people may have shmilar constructs and psychological processes. The sociality corollary states that peeple can play roles in the social field of another through their understanding of the outlook of another person. These corollaries are sufficient to explain social relations in a system that is based upon individualized, personal constructs. Kelly determined that ... by recognizing the subsuming of other people's construing efforts as the basis for social interaction, we have said that social psychology must be a psychology of interpersonal under- standings, not merely a psychology of common understandings (p. 209). Kelly employs a restricted definition of a role as a "psycho— logical process based upon the role player's construction of aspects of the construction system of those with who he attempts to join in a social enterprise" (p. 97). A role is anchored to an individual's construct system and not to situational attributes as in most socio- logical concepts of the term. Role constructs are "constructs which have other persons as elements in their context.... They are constructs which have the presumed constructs of other persons as elements in their context" (p. 209). The postulates of personal constructs and constructive alterna- tivism were developed in conjunction with Kelly's proposal "to postu- late a process as the point of departure for the formation of a psycho- logical theory" (p. 37). Osgood, Suci, and Tannenbaum (1957) also described a theory of perception based upon cognitive attributes of the perceiver. A dimensional model is applied by both Kelly (1955) and Osgood g£_al, (1957). A "construct", in Kelly's theory, is assumed to be a bipolar dimension for construing similarity/dissimilarity. Osgood g; 31. (1957), describes three specific factors in a dimensional repre- sentation. Characteristics of the dimensional model are briefly reviewed. Attributes are represented as continuous, bipolar variables in a dimensional model. Zajoc (Thompson, 1977), describes psychological dimensions in the following way: A psychological dimension is one's capacity to map consistently a set of responses onto a collection of stimuli that is itself ordered. A specific act of 'perceiving' or 'cognizing' a given stimulus object or event is regarded as involving the projec- tion of a stimulus onto a set of psychological dimensions, and thereby attributing to it one value from each of these dimen- sions. These projected values, attributes, are the elements of the cognitive structure under analysis. They are what is commonly understood by the traits, characteristics, qualities, etc., of the object, event, or concept as the person perceive them. (p. 27). Accordingly, attributes are treated as vectors in a n—dimensional space. Factor analysis and multidimensional scaling techniques are used to interpret data in terms of factors, or dimensions with assigned weights. The concept of psychological distance is synonymous with but broader than perceived similarity. Psychophysics attempts to bridge psychological distance with Euclidean distance. Equating psychological distance with euclidean distance has two advantages according to Jackson (1969). The method provides generality, allowing for inferences on many social psychological attributes, and has the advantage of being analytic, permitting geometric and matrix algebra transformations. This model has been accurate in studies on perception of physical stimuli (Jackson, 1969), however, in the domain of personality theory and social processes the lack of criteria for goodness of fit often generates difficulties. Osgood ggnal. (1957), assumed a single set of factors that are embedded in the cognitive components of semantic ability. These dimen- sions are evaluation, potency, and activity. Factor analytic studies by Osgood g£_al, (1957) support this assumption. The three dimensions were found to be independent when describing impersonal events. Potency and activity became less independent when describing people and social events. Thompson (1977) discusses the assumption made by Osgood 35 $1., that a single set of attribute dimensions applies to all cognitive ele- ments. This is a restricted dimensional model. Another model of attribute structure can be postulated. Thompson (1977) states ... such a model-a domain-specific dimensional mode1--suggests that cognitive elements may be divided into categories on the basis of some set of characteristics, with possibly differing sets of attribute dimensions being used within each category (p. 28). The domain specific dimensional model allows greater flexibility than the model used by Osgood £5 31. (1957). The theories of Kelly (1955) and Osgood (1957) describing per- ceptual processes have been reviewed. An emphasis on perception and the mutual phenomena involved in a social relation was traced to the Gestalt school of psychology. The dimensional model of attribute representation was also reviewed. The dimensional model underlies the use of a multidimensional scaling analysis in the study as well as in the theories of interpersonal behavior and personality that will now be discussed. Attributes of the people being perceived are the second relevant area in interpersonal perception. Theories of interpersonal behavior describe and categorize these attributes. Theoretical foundations in the field of personality have a broad empirical base. Typically, representations used to describe interpersonal behavior have been two dimensional. Earliest findings are the observations of Hippocrities (Adams, 1964). Sanguine, melancholic, choleric, and phlegmatic tempera- ments reveal behavior that may be categorized above or below the mean on two dimensions. Foa (1961) found a "strong convergence" of concep- tualization in two dimensional schemes developed from many different lines of research. For example, Leary (1957) devised a system for classifying interpersonal behavior on two scales, love-hate, and dominance-submission. Adams (1964) describes the two axes. The affection-hostility dimension reflects variations in the degree of positive or negative affect manifested toward others. The positive extreme describes warm, friendly, kind, affiliative acts, while the negative extreme describes hostile, critical, angry, disaffiliative acts (p. 195). One pole of the dominance-submission axis is defined by acts of self-confident, assertive leadership and achievement in the face of obstacles. At the opposite pole are acts of passivity, sub- mission, and acquiescence (Po 195)- Another researcher, Hurley (1975), also highlights convergent evidence identifying two dimensions of interpersonal behavior, but leaves open the possibility of other dimensions emerging. He labels the two dimensions as self acceptance-rejection, corresponding to a 10 dominance-submission dimension, and acceptance-rejection of others, corresponding to an affection—hostility dimension. Although these two dimensions appear firmly entrenched, the question of new dimensions accounting for additional variance continues. Hurley (1975) concludes, "other important interpersonal dimensions will probably emerge from future research, but their chances of eclipsing acceptance-rejection of others, and self acceptance~rejection do not seem large." Other theorists have not been contented with a two dimensional framework. Bales (1950, 1970) has developed a three dimensional eucli- dian representation of interpersonal behavior and group social structure. The three dimensions are labelled likeability, control, and task orien- tation. Bales (1970) states that "it is desirable to go further than two dimensions when one is interested in the relation of values or attitudes to interpersonal behavior" (p. 52). The task orientation factor is a values dimension. The poles represent convergent versus divergent values and task orientation. Likeability and control corre- spond with the two dimensions noted earlier from Leary (1957) and Hurley (1975). All of the above theories of interpersonal behavior are subject to two lines of criticism. First, the dimensions typically are arbitrarily selected and represent the investigator's §_priori decisions in research methodology. Second, individual differences in the subject's social perceptions are ignored. This hinders the power and generality of these theories in providing a comprehensive explanation of interpersonal behavior. 11 A third theoretical area pertinent to interpersonal perception involves attributes of the situation. The dynamic system of small groups is of particular importance in the present study since data are collected from small interpersonal process-oriented groups. Descriptions of stages of group development have been reviewed by Tuckman (1965), and Cohen and Smith (1976). A five stage model of group development inte- grates information from the work of both theorists previously mentioned. A trend toward more meaningful perceptions and understanding of others evolves through the phases of group development. The first stage of group development involves testing and depend- ence. The participants' interactions are characterized by expressions of dependency directed toward group leaders and other participants. Frequently, quick solutions or structures are set up to deal with a situation in which norms are not yet specified. Inclusion and belong- ing are central issues of participants. Perceptions of others often rely on stereotypes since participants are primarily in the process of getting to know one another. Cohen and Smith (1976) describe this as a "superficial acquaintance process" involving "categorizing one another or ... pigeon-holing each other with outside roles and statuses often determining inside ones" (p. 76). Tuckman concludes that "orientation, testing, and dependency constitute the group process of forming" (p. 396). The second stage of group development is characterized by intra-' group conflict. As Tuckman (1965) describes, "group members become hostile toward one another and toward a therapist or trainer as a means 12 of expressing their individuality and resisting the formation of group structure" (p. 386). Issues over control result in polarization and competition. Cleavage between dependent and counter-dependent factions within the group may emerge. Cohen and Smith (1976) identify primary group transferences as occurring at this stage. Participants behavior and perceptions are clouded within the turmoil of an ambiguous situation. Cohen and Smith (1976) suggest that this ambiguity ... may lead group members to respond to the leader (or other members) with feelings and behaviors learned in earlier, usually family, and other primary group relationships, i.e., they tend to transfer primary feelings and modes of relating (p. 83). Tuckman (1965) labels this stage as "storming". A third stage describes the development of group cohension. New norms and roles are adopted. Self-disclosure becomes more personal. A sense of harmony develops as participants become more accepting of each other and group leaders. The emphases on commonality and mutuality may affect social perception. Honesty may be less important than harmony. Cohen and Smith (1976) state that "Individual identity is sub- merged in the group in that members deny their own identity in pursuit of group unity" (pp. 91-92). Tuckman (1965) refers to this as the stage of "norming". Functional role relatedness is identified in the fourth stage. The group process is more collaborative. Participants are simultaneous— ly autonomous and cooperative. Self disclosure becomes more open as roles become flexible and functional. Cohen and Smith (1976) describe an increase in the development of empathy and also state that: 13 there are more unbiased evaluations of the contributions of members, and members' questions are evaluated with less regard for power or status in the group. Insight into others is common with group members perceiving defenses, faulty value systems of other group members, and seeking to understand underlying reasons (p. 96). Tuckman (1965) labels this stage as "performing". The final stage of the group is generated by the final task of separation and termination of the group. Tuckman (1965) has ignored this stage completely. Cohen and Smith state that participants' behavior varies. Overly optimistic evaluations about the effects of the group, withdrawal from group interaction, or denial may all be responses of certain participants. Separation may evoke sadness and regret as well as excitement about completion and implementing new discoveries. The process of termination is the least well described stage of development. The dynamic system of an interpersonal process group can be expected to influence social perceptions within the group. The complex- ity, nature, and salience of dimensions of social perception may shift as the developmental trend toward more meaningful perceptions and under- standing of others emerges in the group. Three areas of psychology have been identified as relevant to social perception and have been reviewed. These areas include: (a) theories of perception, (b) theories of interpersonal behavior and personality, and (c) theories of small group development. An integrative focus of these three areas is needed to approach the complex field of social perception. 14 Overview A review of the relevant literature on social perception in groups is presented in Chapter Two. The design of the study is described in Chapter Three and includes detailed information regarding the sample, measures, research questions, and means of analysis. The analysis of results is contained in Chapter Four. The study will contribute informa— tion in the area of interpersonal perception through analyzing the rele- vant dimensions underlying the interpersonal perceptions of members of a small group. CHAPTER 11 REVIEW OF THE LITERATURE Multidimensional scaling techniques have aided the study of the relevant dimensions in an individual's internal representation of others. Studies employing multidimensional scaling of interpersonal perceptions are of concern to the present study. Three studies using multidimen- sional scaling of interpersonal perceptions are reviewed in depth in this chapter. The first study examined fraternity brothers' perceptions of one anothers' personality. The second study analyzed perceptions of members of an ongoing work group. The last study described is a multi- dimensional scaling analysis of T-group members interpersonal percep- tions. This chapter reviews research on the multidimensional represen- tations of interpersonal perceptions in a variety of groups. One of the earliest multidimensional scaling analysis of social perceptions was reported by Jackson, Messick, and Solley (1957). The research was designed to investigate the usefulness of multidimensional scaling for structuring the perception of personality. The investigators were seeking to overcome the problem of choosing §_priori dimensions. They state that "g.priori abstraction of reasonable dimensions in rela- tively unexplored areas, may be too many, too few, or generally irrele- vant for adequately describing a particular domain" (p. 311). The multi- dimensional scaling method of successive intervals was used to determine 15 16 the number and structuring of dimensions of perceived personality. The subjects, who served as stimuli and judges, were 20 college age males who were members of a social fraternity and lived in the same house. Each subject was presented with a randomized list of all pos— sible pairs of the 20 names. They were asked to judge the distance between each person and every other person with respect to similarity in personality along a nine point scale from very similar to very dif- ferent. The average judged distances were obtained for input into a multidimensional scaling analysis. The twenty subjects completed the Stern's Activity Index in a group administration after the collection of the similarity data. The Stern's Activity Index is a 300 item in- ventory assessing the extent to which a person likes or dislikes par- ticular types of activities. The inventory is theoretically linked with Murray's need theory (1938). Friendships ratings, ACE intelligence scores, and age were also obtained. The multidimensional scaling analysis solved for a F—matrix with a rank of.4 rotated to approximate simple structure. The rank of the matrix corresponded to the number of dimensions. The first two factors accounted for over 701 of the variance in ratings while the remaining factors were successively smaller. Factor 1 was best described by the theoretical-intellectual area of the Activity Index with a correlation of -0.42. Factor 2 correlated 0.75 with the friendship ratings. Factor 3 correlated -0.46 with age and was interpreted as a status di- mension since age corresponds with class status and most likely time as a member of the fraternity. Factor 4 accounted for residual variance 17 and was not interpretable. The interpretations of the dimensions are considered tentative and suggestive. The investigators state that more reliable information is needed for an external criteria to be used in labeling the dimensions. Jackson, Messick, and Solley (1957) conclude that "The results suggest that the method is appropriate and extremely promising for the study of social perception. Judgments about 20 unique and complex per— sonalities were found to be ordered parsimoniously along not 20, but a relatively small number of dimensions" (p. 316). A lack of correspond- ence was found between the Activities Index and the dimensions important to fraternity brothers' interpersonal perceptions. This implies limited congruence between the psychologists' and laymens' terms. The inves- tigators suggest factors such as personality style and structure may be important. In summation, the study by Jackson, Messick, and Solley (1957) demonstrates the suitability of multidimensional scaling in exploring social perception. One weakness of the design is the successive inter- vals model which requires averaging ratings across group members to define a group space. This prohibits the analysis of individual differ- ences in the dimensions of perception. A different method of multi- dimensional scaling is required to perform.an accurate evaluation of the individual, as well as group representations. Jones and Young (1972), in a longitudinal study of an ongoing work group, describe the application of an individual differences multidi- mensional scaling analysis (INDSCAL) (Carroll and Chang, 1970). l8 Similarity ratings of all possible pairs of group members are made by each subject. The method inputs each similarity matrix constructed from the similarity ratings made by each subject. INDSCAL reports the amount of variance accounted for by an n-dimensional representation of the similarity data. The INDSCAL method yields two types of configura- tions, a group stimulus space and a subject space. The individual sub- ject space is related to the group stimulus space by individual subject weights reported for each dimension. The subject weights matrix indi— cates the salience of the dimensions to that subject. This technique allows an analysis of individual differences and consequently is more powerful than other multidimensional scaling techniques. The INDSCAL model is used by Jones and Young (1972) to answer four questions about the perceived social structure of a work group: (a) What attributes or dimensions do individuals use in judging or perceiving others? (b) Are subject weights on the dimensions related to role differences among the individuals? (c) What is the social structure of the group and can it be used to predict interpersonal behavior? (d) How stable is the social structure over time? The study used students, faculty, and staff of a psychometric laboratory as members of a long-term intact social structure. Data was collected in February 1969, and February 1970. The 17 stimuli were a subset of 19 judges in 1969 and 31 judges in 1970. This setting was selected for reasons of convenience, the physical proximity of the work group, and the well-defined role and status differences. 19 A number of instruments were administered. The subjects rated their familiarity with the stimulus person. They rated the similarity of all possible pairs of stimuli persons. Subjects rated each stimulus person on several unidimensional scales indicating perceived degree of interest in professional activities. Finally, subjects were asked to choose the two groups of three individuals with whom they associated least and most frequently for advice on research, and socially. INDSCAL analysis of the 1969 and 1970 data were run. Three derived dimensions were labeled through correlations with the unidimensional rating scales. Dimension 1 was labeled "status". Projections onto this dimension correlated over 0.90 with mean unidimensional ratings of status. Dimension 2 was labeled ”political persuasion". Scale values on this dimension correlated over 0.80 with a left-right or liberal- conservative political spectrum. "Professional interests" was the label for Dimension 3. This scale correlated high with interest in statis— tical problems (-0.9l), interest in content areas (0.72), and interest in experimentation (0.85). The 1970 results replicated the 1969 results in most cases. The correlations between the 1969 and 1970 status, political persuasion, and professional interests dimensions were 0.92, ~0.94, and -0.92, respectively, indicating that the structure of the group did not change very much. Faculty and graduate student judges differed in how they used the dimensions. Faculty relied most heavily on status. Graduate students tended to weigh political persuasion and professional interests more heavily. The salience of the dimensions shifted systematically from 20 one year to the next. Jones and Young (1972) state that "The overall pattern of these shifts in salience suggests that the stability of a subjects' perception of the relationship between the various members of the group, including himself, increased as a function of time in the group" (p. 118). The authors were able to successfully use the three dimensional group space to predict the formation of doctoral committees, the fre- quency of interaction between individuals in social contexts, and the frequency of interaction for research collaboration and advice. The relationship between sociometric choice and perceived group structure was very strong. The authors interpret the sociometric choice results in terms of social comparison theory (Festinger, 1954) which assumes that individuals are most attracted to people perceived as similar to themselves. The study by Jones and Young (1972) demonstrates the suitability of the INDSCAL model for studying groups. The model allows the deriva- tion of the "relevant" dimensions of similarity that underly sociometric choices. Methods making §_priori selections of variables must deal with the fact that many variables and selection criteria are available to choose from. Many different representations of the same social struc- ture can result from these choices. The INDSCAL model bypasses this problem by reporting the dimensions relevant to the subjects perceptions. The study demonstrates the accuracy of predictions based upon a social structure represented through subjects' perceptions. 21 The three labels for the derived dimensions in the report of Jones and Young (1972) are of interest to the present study. The dimensions of interpersonal perception were "few and interpretable" (p. 119) as hypothesized. Status may be interpreted as involving power and influ- ence in the system. The political values dimension corresponds to specific values and beliefs. The third dimension is descriptive of professional interests. The effective functioning of the work group must involve an awareness of these professional interest areas. The authors note the absence of a like-dislike dimension in the social struc- ture. They suggest that the affective dimension "is an important determinant of interaction only when role differentiation in a group is not too great" (p. 119). This would imply that an affective dimension would be a factor in the less structured groups described in the present study. Thus, Jones and Young provide some preliminary evidence identi- fying the dimensions of interpersonal perception. A study to describe the "nature and development" (p. 44) of T-group members social perceptions during the life of a group has been reported by Lewis, Lissitz, and Jones (1975). Questions were raised in four areas including "the nature of dimensions along which members per- ceive one another, the complexity of their perceptions, the stability of interpersonal perceptions over time, and the relative importance of the interpersonal dimensions used at various points in the development of the group" (p. 44). The investigators assume the underlying dimen— sions of perception are interpersonal. They also state that "As members of the group gain an increasing amount of direct experience with one 22 another, their perceptions of one another might be expected to become more differentiated" (p. 44). The T-group that was studied met for 13 ninety minute sessions over six and one-half weeks. The subjects were five undergraduate students who volunteered to participate. The sixth subject, and group leader was a male clinical psychologist. "Similarity ratings were ob- tained immediately after each group meeting with the use of a questioneer that called for similarity judgments of all possible paired comparisons among the six members of the group" (p. 45). The leader rated each group member after each session on the eight following bipolar scales: popular-unpopular, uninvolved-involved, traditional-radical, empathic- insensitive, defensive-open, active-passive, anxious-calm, and weak- strong. "These eight rating scales were chosen to represent a range of important interpersonal characteristics and are similar to those described by Carson (1969) in his survey of interpersonal behavior" (pp. 45-46). The study reports a three-dimensional solution had a "substantial- ly higher correlation coefficient with the actual similarity ratings (0.73) than did the two- or one-dimensional solutions (0.65 and 0.52, respectively)" (p. 46). The correlations between the three dimensional solution and the similarity ratings improved over the thirteen group sessions indicating an increasing tendency for group members to view one another along the same three dimensions. Interpersonal perceptions neither increased nor decreased in complexity over the thirteen sessions. 23 The INDSCAL dimensions were labeled by Lewis, Lissitz, and Jones through their relationship with the group leader's average post-session ratings. A multiple regression analysis was performed. A simple product-moment correlation of -0.81 was reported between the second INDSCAL dimension and the traditional values--radical values rating scale. The first INDSCAL dimension was not as well-described yielding a 0.41 correlation with the same traditional values-radical values dimension. A multiple correlation of 0.50 resulted with the addition of the popular-unpopular rating scale. The third INDSCAL dimension had a multiple correlation of 0.51 with a scale along an actively calm to passively anxious continuum. A matrix of subject weights was presented to determine the importance of each derived dimension to each group members' perceptions. The study reported that "the response patterns of the T-group leader and the other group members differ markedly. The group leader made considerably greater use of the second INDSCAL dimension, traditional- radical, than did any other group member and considerably less use of the first" (pp. 46-47). The fact that only one group leader was studied limits generalizations about the leader role. The greatest strength in the design used by Lewis, Lissitz, and Jones (1975) is in the collection of data after each session. The come plex nature of members' perceptions was found to have meaningful representation in three dimensions, and the dimensions were found to be relatively stable throughout the group life. Repeated measures after each group session also allowed for comparison of subject weight shifts 24 with the proposed developmental processes of the group. For example, it was reported that dimension 3 (actively calm to passively anxious) was more important for making judgments earlier in the group than it was in later sessions. The authors see this as consistent with Tuckman's (1965) theoretical framework describing the forming and storming periods of group development in which anxiety is very high due to the unstructured situation. The study presents a potent approach to in- vestigate group development. One failure of the study was to provide an adequate label for the dimensions. The average ratings of the group leader did not corre- spond to all three INDSCAL dimensions. The authors pointed out two problems with the procedure. First, the ratings which were averaged across the thirteen sessions actually had changed considerably for certain group members. Secondly, the leaders'perceptions were reported as differing from other group members because of a heavily weighting of dimension 2. This may have easily contaminated what was to serve as an external criteria. Other possible methods of analysis may involve tak- ing ratings from all group members, or taking ratings from observers. A third criticism of the procedure may be leveled at the selection of bipolar scales. The scales were chosen to represent a range of impor- tant interpersonal characteristics; however, there are no theoretical supports for these choices. Attempts at labeling derived dimensions may be more successful by attending to the theories of interpersonal behavior in creating any rating scales to serve as an external criteria. 25 Summary Three studies using multidimensional scaling techniques on social perceptions have been described. Each study exhibits the usefulness of multidimensional scaling in the field of social perception. The ability to obtain ratings unbiased by an investigator's §_priori choices of important variables is one distinct advantage of multidimensional scal- ing. INDSCAL (Carroll and Chang, 1970) has the additional advantage of indicating the quality and degree of individual differences in social perception, thus adding to the robustness of the analysis. The parsimony and power of the INDSCAL technique in the area of interpersonal percep- tion have substantial implications for future research. All of the research reviewed indicate that the dimensions under- lying social perceptions are relatively few. Jackson.g£_gl, (1957) dis- covered a 4 dimensional representation. Jones and Young (1972) found a 3 dimensional solution, and Lewis 55.51, (1975) described a 3 dimen- sional solution. Complex judgments were found to be structured parsi- moniously along a relatively small number of dimensions (see Table 2.1). The interpretability of the dimensions of perceptions is a diffi- cult problem. Interpretation is complicated by the need for a reliable external criteria. Jackson ggngl, (1957) used activities reports, friendship ratings, and age as variables to label four dimensions under- lying fraternity brothers perceptions of one another. Dimension 1 cor- related -0.42 with theoretical-intellectual activities. Dimension 2 correlated 0.75 with friendship ratings. Dimension 3 correlated -0.46 26 with age and Dimension 4 was not interpretable. Jones and Young (1975) used unidimensional scales to assess interest in a variety of social and work areas of an ongoing academic work group. Status correlated above 0.90 with dimension 1. Political persuasion correlated above 0.80 with dimension 2. Professional interests correlated highly with dimension 3. Lewis ggngl, (1975) reported correlations of 0.41 with a traditional-radical scale -0.81 with a traditional-radical scale, and 0.47 with an active-passive scale for the three respective dimensions used by members of a T-group (see Table 2.1). The study by Jones and Young (1972) was most successful at interpreting the dimensions. The two other studies were less successful. The differing social struc- tures may be a cause of this. The work group has well-defined roles. The labels found by Jones and Young (1972) correspond to the group roles. Political persuasion is an exception to this since it describes more personal characteristics. Judgments in the fraternity and T-group involve more personal characteristics. Jackson g£_§l, (1957) suggest variables such as socioeconomic status and social skills may provide a better fitting criteria. These factors appear to be far too global. A specific evaluation of the type of interpersonal behavior exhibited would lead to clearer labels, particularly in a group involving a degree of personal intimacy between the members. The research by Lewis 35 El: (1975) evaluated the members interpersonal behavior. This design was hindered by the lack of a coherent theory of interpersonal behavior and by reliance solely upon the leaders ratings. An accurate means of interpreting the dimensions of a multidimensional scaling analysis can occur only through reliable external criteria. 27 .mnaa .oosoh one nuanoAAgowsvo .Nuaa .ussow one meson A .aan .eueeom can scene»: .aomsomna uaeououmuouoa uoa e newshound o>fiooeelo>fiuoo no.0 Hoooamoomouo nw.o o>ooe Aosueuov owe oe.ou m Hooaomu soaooso nemaoeoeeeuu Hm.ou name Hmoeueeoa om.o u>onm museum“ eeeaeaoeuu ma.o a :5... .3... . . 3:2... smegma“ . H Iaeooauaoeuu He.o ca o e H H a Na 0 Huang .uuou Henna .uuoo Huang .uuoo does onsoHMIH oqoouu duos ohufiouououh Inoafia wdaaoom HoooHosoaHowuHszflomsoueH oowmausoon sowuaoouom HesoouoeuouaH we ooowosoawa .H.~ «Home 28 Few studies examining the interpersonal perceptions between people in a group have been reported. Each of the three studies reviewed in this chapter had one group as the unit of analysis. Limited generaliza— tion can be made from these studies in spite of assumptions about the similarities of these groups with other groups of their particular class. Further research is needed to broaden this base of analysis of social perceptions in groups. The implications of these three studies have been integrated into the design of the present study. This is described in the next chapter. CHAPTER III DESIGN OF THE STUDY Information describing the population sample, measures, research questions, and analysis of data in the study is contained in the present chapter. Sample Data were collected from students enrolled in two undergraduate courses offered at Michigan State University during Winter term, 1977. One course was offered through the Department of Psychology and the other through the College of Education. The courses were concerned with group processes and interpersonal communication strategies. Ongoing small groups were used as interpersonal behavior laboratories. Three groups from the psychology class and two groups from the educa- tion class participated in the study in exchange for individual and group feedback given at a later date. PsychologylGroups The three groups affiliated with the psychology class lasted for nine weeks. The groups met for two ninety-minute sessions each week with two twelve-hour extended sessions during the third and seventh 29 30 weeks of the term. Course grades were based partially upon group attendance. Consequently, few sessions were missed by group partici- pants. Group participants were junior- and senior-level undergraduate students who were screened by the course instructor prior to enrollment. Participants self-selected a group at the start of the term, most often on the basis of scheduling convenience. An effort was made by the instructor to make the groups balanced in sexual makeup. Group members were instructed about the confidential nature of group material and were requested to limit contact with one another outside of the group. The class, which included all three groups, met once a week with the instructor to cover didactic material on interpersonal behavior. Group leaders were selected and trained by the course instructor. Leaders were graduate and undergraduate students who had to have experi- ence as participants and, in some cases, as observers of a group. Also they must have demonstrated competent interpersonal communication in their group experiences. All group leaders met with the course instruc- tor and other leaders once a week for supervision. The group process was primarily "here and now" interaction oriented. Group leaders typically provided minimal structure to the group's activities. They were trained to make group process interventions as well as personal interventions. For the convenience of the study, group leader and group participant are defined as subsets of the term group member. Consequently, throughout the study, the term group member is by defini- tion inclusive of leaders and participants. 31 The psychology groups were labeled group A, B, and C. Group A had eight members; five were males and three were females. There were two co-leaders, one male and one female. Group B consisted of eight members; five being male and three being female. Group B had three co-leaders; two of the leaders were male and one was a female. Group C also had eight group members. It was equally balanced with four males and four females; however, the two co-leaders were both females. One female participant in group A was asked to leave the group just prior to data collection; however, data from this participant were in- cluded in the study. The course instructor had each group member complete ratings of all group members, including themselves, on a series of semantic dif— ferential scales during the fifth and ninth weeks of the groups. These ratings scales clustered along two dimensions identifying self accept- ance-rejection and acceptance-rejection of other. A self report and others report were returned to the group members as feedback. An administration of these rating forms was done three weeks prior to the data collection of this study. Education Groups Data were also collected from groups in an undergraduate education course (Education 200). The course is a required introductory class for all education majors and is designed to train students in communication skills necessary for effective teaching. Reading materials, lectures, individualized carrel activities, and tutorial services are used to develop skills in the content demands of teaching. The Interpersonal 32 Process Lab (IPL) facilitates the development of skills in group process and interpersonal communication strategies. The structure of an IPL group is distinctly different from an encounter or sensitivity group. The purpose of an IPL group is to teach specific interpersonal skills. The IPL leader will lecture, initiate exercises in communication skills, and monitor the group process for self-disclosure and the communication of responsible feed- back by group members. Students must demonstrate competence in inter- personal communication skills in the IPL section to pass the course, which is graded pass-no pass. Students who receive non-passing grades are allowed to participate in an IPL group another term. The group leader is responsible for evaluating the student's skills in the IPL group. Lopis (1975) describes the specific skills that are evaluated: (a) Interaction Assessment—-"The ability to identify the cues of an interaction as primarily affective (dealing with feelings), or cognitive (dealing with content) and to respond appro- priately." (b) Active Listeningr-"The listener takes an active responsibility in helping the speaker's self-clarification." (c) Exploratory Questioniggf—"The ability to seek further informa- tion or clarification for self and others without cuing a particular response." (d) Observation Skill-—"The ability to recognize and interpret diverse modes of nonverbal communication; that is, hands, face, posture, gestures, and so on." (e) Self Disclosure-~"The ability to share one's own ideas, opinions, and feelings as they relate to the immediate situ- ation." (f) Feedback Skill--"The ability to relate honest reactions caused by another person's behavior to the first person. A rationale is included to determine whether appropriate reaction would be positive or negative and to distinguish responsible reactions from irresponsible reactions." 33 (g) Value Awareness--"The ability to recognize, for self and others, consistency or inconsistency between one's professed value system and the expressed values interpreted through behaviors." (h) Process Model--”The ability to systematically assess behavior, set goals, identify strategies, and to evaluate behavioral changes in relation to the stated goals." Typically the leader will introduce these skills to the group at a graduated rate with the most structure occurring early in the term. Structure is reduced after all concepts have been introduced allowing students time to initiate process and demonstrate the skills. Students are given a feedback sheet twice during the term on which the leader evaluates their demonstrations of the skills. Students in the IPL groups were typically in the second or third year of their undergraduate program. They selected the particular IPL section from a pool of over thirty sections on the basis of convenience in time and location. They met twice a week with approximately five hours per week in the group. Group leaders were graduate students in education and met weekly as a staff for training and supervision from the course instructors. The leaders were selected by the instruc- tors and employed by the university as teaching assistants for the dura- tion of the school year. The two smallest IPL sections were selected to participate in the study so that the Education and Psychology groups would be similar in size. The IPL groups are labeled group D and E. Group D consisted of nine group members including two male and six female participants. The leader of group D was male. Group E had eight members composed of two male and five female participants and had a male leader. 34 Measures Five sets of rating booklets were constructed that contained simi- larity rating forms and semantic differential rating forms. The names in each set corresponded with the members of either group A, B, C, D, or E. Subjects rated only members of their own group. Groups A, B, and C were administered the instruments during the weekly class meeting that included all three groups. Groups D and E completed the instruments independently at their regularly scheduled group time. Two rating tasks were included in the booklet. The first task consisted of rating on a Likert scale all pairs of stimulus persons in the group. The second task was a group semantic differential rating of all group members. Data were collected during the eighth week of the ninedweek term. A seven-point Likert scale was used for both ratings scales and was selected on the basis of the following research. Miller (1956), after a review of research on the absolute judgment of simple unidimen- sional stimuli, suggested that the average person's capacity for process- ing information is limited in accuracy with a span of absolute judgment in the neighborhood of seven. Green and Rao (1970) report that limiting response categories to two or three deteriorates information recovery, and they also found that product moment correlations increase with the fineness of response categories up to six. Another researcher, Finn (1972) concludes that, when "taking into account reliability of ratings and the desire to maximize variances of ratings" (p. 264), the optimal level is six or seven response categories. Evidence that testing time increases with an increase in the number of response categories and 35 that neutral responses increase with a three- or five-point scale as opposed to a seven- to nineteen-point scale has been reported by Matell and Jacoby (1972). The research indicates using seven response cate- gories maximizes the variance of the ratings without presenting an excessive number of categories that may exceed the subject's ability to discriminate; hence, the similarity ratings and group semantic differ- ential ratings utilized seven-point scales. A sample form of the rating booklet is contained in Appendix A. Similarity Ratiggg The first two pages of the rating booklet contained the similarity ratings task. Subjects were instructed to rate the similarity of the group members according to what they felt to be the most relevant attributes of the individuals. Each group member rated all the possible pairs of individuals, including themselves, within that member's group. A seven-point Likert scale with the rating of one labeled very dissimié lar and the rating of seven labeled very similar followed the names of each pair of group members. The list of pairs of group members was presented in a Ross (1934) ordering. This method yields a balanced order of pairs that Ross (1934) states has the following advantages: "(1) They maintain the greatest possible spacing between pairs involving identical members, (2) they are so balanced as to remove time and space errors, (3) they avoid regular repetitions which might have suggestion effects" (p. 382). The Ross ordering facilitates the rater in treating each pair to be rated independently of the preceding pairs. 36 These similarity ratings were used to construct the half matrices that are used as the input for the multidimensional scaling analysis. Group Semantic Differential (GSD) The GSD consisted of sixteen sets of bipolar adjectives and followed the similarity ratings in the booklet. Subjects were in- structed to describe each group member on the adjective scales and to be as discriminating as possible by using the entire scale. A bipolar adjective headed a seven-point Likert scale that followed each name on a list of group members. The polarity direction of the sixteen GSD ad- jectives was randomly assigned to control response sets. The listing of group members names maintained the same order throughout the sixteen ad- jective pairs. Scores were summed across group members; providing an average group rating of each group member on each bipolar adjective. The GSD provided a means of structuring the perceptual domain of the group members. The structure described by the GSD was compared to the multidimensional scaling analysis in an effort to label the dimen- sions underlying the similarity ratings. The selection of bipolar adjectives was crucial to the success or failure of the descriptive function of the GSD. Maguire (1973) described four criteria in select- ing scales for the semantic differential: (a) They must be representa- tive and provide adequate coverage of the domain, (b) They should be well-defined for the population of interest, (c) They must be appro- priate as descriptors of the stimulus of the study, and (d) They must be polar opposites. The literature in the area of interpersonal behavior in groups was combed for descriptive words and phrases 37 to select an appropriate and broad range of adjective pairs. The four factors selected were dominance, affiliation, activity, and goal orien- tation. Four sets of bipolar adjectives were selected to represent each factor. Strong-weak, independent~dependent, leads—follows, and rugged-delicate were expected to describe the dominance scale. Accepting-rejecting, warm—cold, friendly-unfriendly, and gentle-harsh were used for the affiliation scale. The activity scale was described by active-passive, fast-slow, talkative-silent, and involveddwithdrawn. Goal orientation included orthodox-heretical, loyal-disloyal, tradi- tional-radical, and obeys-disobeys. Each person's position on a par- ticular factor was calculated by summing the average scores on each of the four subscales. The order of presentation of the sixteen bipolar adjectives was randomly assigned. The sixteen bipolar adjectives selected for the GSD were taken primarily from research reported by Burke and Bennis (1961), Hartley (1968), Hurley (1976), Osgood g£_§l, (1957), and Bales (1970). Burke and Bennis (1961) reported a factor analysis of a group semantic dif- ferential instrument with nineteen bipolar adjectives. Three factors accounted for 86 percent of the variance in ratings. Factor A was labeled evaluation and had high correlations with nine of the adjective pairs. The bipolar adjectives and the correlations were: friendly- unfriendly (0.90), accepted-rejected (0.90), adaptable-rigid (0.87), harmonious-discordant (0.87), good-bad (0.84), warm-cold (0.81), sensitive-insensitive (0.77), close—distant (0.79), and included- excluded (0.74). Factor B was labeled potency and had high loadings on the following six adjective pairs: strong-weak (0.93), 38 independent—dependent (0.91), leads-follows (0.84), important-unimportant (0.74), hard-soft (0.72), and central-peripheral (0.71). Factor C had high loadings on three adjective pairs. These three bipolar adjectives were talkative-silent (0.85), involved-withdrawn (0.78), and active— passive (0.75). Factor C was labeled participation. Another research; er, Hartley (1968), used a semantic differential for assessing group process changes. Evaluative, potency, and activity were the three factors measured. The evaluative factor was composed of cruel-kind, unfriendly-friendly, tense-relaxed, annoying—pleasing, and fair-unfair adjectives. The five adjective pairs describing potency were shallow- deep, unconcerned—concerned, easygoing-aggressive, soft-loud, and weak- strong. The third factor activity, was derived from quiet-talkative, repetitive-varied, remote—intimate, insensitive-sensitive, and passive- active adjectives. Hurley (1976) used the semantic differential to assess two factors that are labeled self acceptance-rejection (SAR), and acceptance-rejection of others (ARO). The six adjectives that measure SAR were shows feelings-hides feelings, active-passive, strong- weak, like self—dislike self, open-guarded, and calm-tense. ARO in- cluded gentle—harsh, accepting-rejecting, listens-tunes out, warmrcold, permissive-controlling, and reaches out-holds back adjectives. The median correlation within the sextets of SAR and ARO was r = 0.54, generally supporting the construct validity of the scales. Hurley states that "Improvement in the selection of the ARO subscales is plainly desirable, however, because the reaches out-holds back and listens-tunes out scales, intended to assess ARO, correlated more 39 strongly with some of the SAR scales than they did with other ARO scales" (p. 186). The fourth factor selected for the GSD, goal orien- tation, was selected from the research of Bales (1970). Bales studied task groups and described interpersonal behavior along a dimension labeled task orientation. This factor is characterized by acceptance of group values at one pole and the rejection of group values at the other. The convergent values pole describes a work orientation that is instrumental, conservative, and analytic, focusing on problem solving. Interpersonal behavior is impersonal and serious since the individual is primarily focused on group goals. The divergent value pole reflects an emphasis on expression and fantasy. Heretical rejection of group norms may be expressed in joking, playing, laughing, or side conversa- tions. The finding by Lewis gt_§l, (1975) that a traditional-radical scale describes a dimension of perception in a T—group supports Bales' construct of task orientation. The adjective pairs selected for the activity, dominance, and affiliation scales have previously established validity in the research of Burke and Bennis (1961), Hartley (1968), and Hurley (1976). The adjective pairs describing goal orientation do not have prior validity established as semantic differential scales. The goal orientation adjective pairs were selected by the researcher from Bales' descriptions of that dimension. The adjective pairs utilized for the four GSD scales attempt to provide adequate coverage for the relevant domain and main- tain the independence of each factor. 40 Preliminary results using the original GSD adjective pair clusters suggested that these clusters were not independent. A factor analysis and an elementary linkage analysis (McGuitty, 1957) were performed pggg th.on the data from the GSD adjective pairs. The correlation matrix of the adjective pairs across the forty-one subjects in the five groups (see Table 3.1) was used in the factor analysis. The elementary linkage analysis used the adjective pair correlation matrix from each group. The factor analysis and linkage analysis are methods of clustering the GSD adjective pairs. The two methods identified clusters that were different from one another, as well as from the original clusters. The VARIMAX rotated factor matrices and proportion of variance for the three- and the four-factor solutions are reported in Table 3.2. Two factors, labeled activity and affiliation, accounted for 882 of the variance in ratings in the four factor solution. Two factors, labeled affiliation and dominance-activity accounted for 932 of the variance in ratings in the three factor solution. The factor clusters derived from the factor analysis are described in Table 3.3. Results from the four factor solution showed factor 1 included the four adjective pairs ex- pected to describe activity with the addition of the leads-follows adjec- tive pair, which was expected to cluster with dominance. Factor 2 was composed of the four affiliation adjective pairs with the addition of loyal-disloyal and obeys-disobeys, which had been expected to describe goal orientation. Factor 3 contained three adjective pairs from the dominance cluster. Factor 4 included orthodox-heretical, and tradition- al-radical subscales describing goal orientation. The three factor 41 mm. om. «e. um. ma. mm. NH. mu. co. oh. mn.o mn.l mo. on. mm. mm.l mm.l mm.l on.l co. so.l ma. «0. mo.: ne.l Hm.l em.l on. me. me. me. we. on. Na. em. as. mN.I mm. co. No.l mH.I mH.I flo>aH HA. unseemum>aumaame en. soamuuomm mm. o>Hmmomlo>fiuo< mo. emummloauoou mm. mausoeumsslmaosofium mm. vacuuauoz mm. wcfluoonoMIwcwueooo< Hm. oumoaaoaloowwom em. maoaaoeumemuo mm. uoooooeoalusoosoooooH xmozlwoouum .oa .nH .ea .mH .NH .HH 0 ,4 0.0. HNMQWGNwO‘ 0H ma «H ma NH HH 0H m w m o n e m N nae u my mesouo HH< mmouo< mufimm o>wuoofio< Qmu mo mdoeumaouuoo .H.m maan 42 Table 3.2. Varimax Rotated Factor Matrices and Proportion of Variance 4 Factor Solution 3 Factor Solution Adjective pairs 1 2 3 4 1 2 3 Strong-Weak .60 .24 .71 -.O2 .17 .93 .02 Independent-Dependent .44 -.12 .82 -.20 -.24 .89 -.13 Leads-Follows .86 .16 .42 —.15 .32 .89 -.24 Rugged-Delicate .10 -.33 .75 —.27 -.53 .61 -.16 Accepting-Rejecting .23 .92 .05 .02 .81 .20 .ll Warm-Cold .46 .82 —.12 .15 .93 .22 .10 Friendly-Unfriendly .52 .79 .09 .00 .83 .42 .00 Gentle-Harsh .02 .82 -.27 .34 .81 -.19 .37 Active—Passive .90 .22 .26 -.09 .45 .79 -.23 Fast-Slow .81 .26 .28 -.07 .43 .75 -.18 Talkative-Silent .93 .22 .14 -.14 .50 .72 -.31 Involved-Detached .86 .42 -.02 -.05 .71 .56 -.22 Orthodox-Heretical -.14 .33 -.09 .86 .32 -.15 .88 Loyal-Disloyal .37 .77 .02 .37 .82 .26 .36 Traditional-Radical -.26 .09 -.26 .90 .15 -.37 .80 Obeys-Disobeys .17 .65 -.23 .52 .74 -.06 .48 Proportion of Variance .54 .34 .07 .05 .57 .36 .07 solution reported factor 1 contained the four original affiliation adjec- tive pairs plus involved-withdrawn, loyal-disloyal, and obeys-disobeys. Factor 2 was composed of the four dominance adjective pairs plus three activity adjective pairs. Two adjective pairs, orthodox-heretical, and traditional-radical correlate with factor 3. ~ divided and merged with either affiliation or dominance clusters in the The activity cluster was three factor solution. The labels activity, affiliation, dominance, and goal orientation were maintained for the four factor solution. The labels used for the three factor solution were affiliation, dominance- activity, and goal orientation. 43 Hmuaommlamcofluevoua Hooauouomlxooosuuo uooaemlo>auexaoa aonIuoom u>aooom|o>fiuo< ouoofiaonlvowwsm wsoaaomlooooa TsooooooouuooooomooaH sooznwsouum «homeownlmhooo Heaoameouflmmoe ssouoeuezuoo>ao>sH emuomloausoc afloaofiumcalmaooofium oHoolsuo3 wcwuoohoMIwowueoou< ooauouoowuo Hoowlm ANAH>auoao>eH uooosoeon mausoauwaalmaoaowum usoHHmlo>wooxHeH HoofiomMIHmooHoHomuH Iuooooooooou vaoolauos soamluoom Hoofiuwuomlxoooouuo xoozlweouum mofiuoofioMIwoaueooo< o>eommmuo>euo< sowumusofiuo HooUIq oucmowaonlm cowuoaafimwfiuuAao>oH uooaamlo>fiuoxama emuomuoauoou havooauwsbumaooowum oueofiaooloowwom uooosuooaluooocuooosH HmoHoMMIHmooauaooHH sonIuoom oaoolauoz Moozwwaouum Hmoauouomlxooonuuo o>wmommlo>auo< moquoonomlmoaueooo< msoaaomlmoeoq P onHHHU¢ onHHuoofio¢ one muouosao uwom o>auooho< oouoomxm .m.m manna 44 An elementary linkage analysis classified the GSD adjective pairs in a typal structure. Linkage is defined as the largest correlation that an1 adjective pair has with any or all of the other adjective pairs. A typal structure is defined by McQuitty (1957) as "one in which every member of a type is more like some other member of that type than he is like any member of any other type"(p. 209). The linkage analysis was performed on the adjective pair correlation matrix of each group because it was suspected that the adjective pairs' association with one another would change from group to group. The lower limit defining a signifi- cant correlation for the linkage analysis was i .71, which will describe about 502 of the variance in the adjective pair ratings. The adjective pair clusters derived from the linkage analysis for each group are reported in Table 3.4. Diagrams of the adjective pairs' typal structure in each group are presented in Figure 3.1. The numbers in the diagrams correspond to the adjective pairs as listed in Table 3.1. Three typal clusters were found in group A. Two of the three types are equally large. One of these clusters (Cluster IA) was come plex with items from the dominance cluster and a sharply defined nega- tive pole formed by goal orientation adjective pairs. The other large cluster in group A (Cluster IIA) contained all activity adjective pairs, and one adjective pair each from dominance and affiliation. The third type (Cluster IIIA) in group A was smaller than the first two and described largely by affiliation adjective pairs with the addition of loyal-disloyal and obeys-disobeys. Cluster IIIA also had weak positive connections with type 11A and with type IA's negative pole. nun—'\ *fif‘n 5‘th fir~:*-< U‘nl.>-.~:< mNKr—xcwq khuucmsmqm .v.m, seems 45 .hufiueaoo oomuo>ou noueofioow n "ouoz vooowuoou oeoowol one .oH unoHleo>Huomeu .HH HohoaoaolaomOH .qH BoamlumMM .oa esmuoeufisloo>ao>cw .NH o>Hmmmanu>fiuom .a emumnnoauoom .w n Houseufiomuulflmoeomu .MHI msoHHomsmomoH .m havouaumoslhaoaoaum .m xoooeuuolfimoauouos .MHI usooaoeoolueuoaoeooefi .N oHOUIauos .o oumoflaooloomwsu .q Mooslwoouum .H wefiuoonoulwefiumooum .m HHH HH H mmueowaolmmoeo .oH Hooofiumesl Hvsuauw n IIII Homoawfiolammoa .QH oaounfihos .o csmuosuezlou>ao>cw .NH wcfiuoonoulwswuoooom .m usoHHmIm>fiummeu .HH oumofiaooloowwsu .q o aonuummm .oa wBOHHomlmoouH .n Hmowomulfimoofiueomeu .mH o>fimwoolo>wuum .m unoccumuolucoocoeooow .N Hmoeuouoelxoooeuuo .mH enameloauoow .w Mooslwoouum .H HH HmaoflufiomuulamofiomM .mHI o>wmmomlo>euom .w xovocuHOIHmoHuouo: .MHI enumeluaucom .m empouawpomee .ee- Hmeoemaelameoa .aa eaeemauwaaneaeameue .A oumoflaooloumwou .q csmuocuwsloo>ao>ofi .NH ofioolaums .o m unoocoaoolucooCoomoow .N unoawmlu>fiumxamu .HH wowuoohuulwcfiumooom .m xooslmoouuw .H zofimlumow .oa mBoHHoulmomoH .n HH camuoeuflsloo>ao>ofi .NH Hmoowuaomuulamofioou .mHl usuaflmlo>fiumxaou .HH xoooeuHOIHmoHuouoe .man zueomflolxueo .oa soamlumew .OH mauaumlnmume .wl Hmzoamfiolamzoa .ea m>wwmmnlo>fluom .m mumofiamoloowwsu .q < oHoolaums .o Afloaufiumoolzavoowum .m ucooeomuoluaoocmmooaw .N wcwuoomoulwowuemoom .m waoHHOMImomoH .m emosnmoouum .H HHH HH H uuumsHu msouo osouo comm pom muoumsao uwmm u>Huoofio< mfimzaoe< uwmxowg mueususoam .q.m canoe 46 whoeomwolmwoeo .oa i HmoeomulamoOHuHoouu .mH m osmuoeufisloo>ao>ow .NH oaooufihms .o HomonHoIHm%0H .qH m ucoHHm|o>Hummeu .HH wowuomfioulmafioeooom .n HoowuouoSIxoooeuuo .mH m aonIummm .oa maoHHoMImoooH .m m emumntoauoow .w _ o>wmmmeto>auoo .m usooooooouuooosooooaw .N powwoulouMUwHoo .ql (e waoomfiumaslmaooowum .n xmoslwaouum .H HH H H noumoao macho oosaeuaoollc.m manna Figure 3.1. Note: 47 g Diagrams of adjective pair typal structures for all groups. Heavy solid lines indicate positive typal structure.. Heavy dashed lines indicate negative typal structure. Light solid lines indicate r Z_.7l correlation. Light dashed lines indicate r.: -.71 correlation. 48 Figure 3.l--continued 50 Figure 3.l--continued 13 ....“ 15 L16174’ \ I'T'l 51 The typal structure of the adjective pairs in group Iiwas domi- nated by one major cluster (Cluster IB) that embraced two five-item types mainly consisting of affiliation and activity items. Two goal orientation adjective pairs were strongly related inversely to the two five-item types and defined a negative pole of the major cluster. A small but distinct secondary cluster (Cluster IIB) was composed of mainly dominance adjectives and also had a slight correspondance to the activity portion of the major cluster in group B. Group C was found to contain a comprehensive major cluster (Cluster 1C) of three directly related types with distinctive facets of affiliation and activity in them. A small secondary cluster (Cluster IIC) was made up of two goal orientation adjective pairs and was weakly related negatively with an activity adjective pair in the major cluster. Group D presented the most complex picture of adjective pair typal structure. Six types were identified that were linked with one another in a circular pattern. The pattern was broken by the researcher to form three clusters used in further analysis. The most central type was made up of four affiliation adjective pairs and was joined with a two-item type containing gentle-harsh and obeys-disobeys adjective pairs. Adjective pair 12 was separated from a three—item activity type and joined with the two types previously noted to form Cluster ID. The two remaining items in the activity cluster and a three—item typed structure containing mostly dominance items joined adjective pair 2 to form Cluster IID. Adjective pair 2 was separated from a two-item type that related strongly to the three—item dominance type. Cluster IIID 52 contained item 4 at the negative pole and a two-item goal orientation type at the positive pole. These two goal orientation items also corre- sponded moderately with the affiliation types in Cluster ID. Group E contained four typal structures that were formed into two completely separate clusters. One major cluster (Cluster IE) con- tained two types the largest of which was laden with dominance, affili- ation, and activity items. The smaller type in Cluster IE was formed by only two items, 2 and 10. The second, smaller cluster (Cluster IIE) was primarily composed of goal orientation items. Item 4 found an in— verse pole of Cluster IIE. The results of the factor and linkage analyses indicated that the groups and subjects interpreted the adjective pairs differently than had been expected, consequently, the most independent clusters of adjective pairs do not correspond in all cases with the expected factor clusters. Goal orientation was most clearly defined by only two adjective pairs, traditional-radical and orthodox-heretical which were inversely related to dominance in groups A and D, inversely related with affiliation and activity in group B, and centered a distinct cluster in groups C and E. The two other items from the §_priori goal orientation factor (loyal-disloyal and obeys-disobeys) were clustered with affiliation in the factor analysis, and generally in the linkage analysis, with the exception of group E. The §_priori adjective pairs describing the activity cluster were highly intercorre- lated through the factor analysis and linkage analysis, however, activity was often bonded with either dominance or affiliation items depending 53 upon the particular group studied. Activity items were linked with both affiliation and dominance items in three of the five groups. The inter- relationships within the §.priori dominance and affiliation items were much more complex than expected. The linkage analysis found that the GSD adjective pairs did not cluster into four independent factors in any group. The adjective pairs formed three clusters in two groups and two clusters in three groups. The factor and linkage analyses suggested that the §_priori adjective pair clusters were not the most valid combi- nations for use as an external criteria in the study. The analysis of data initially used the §_2£$2£$,adjective pair clusters. ‘Pg§£_hgg_ analyses were performed using the clusters described by the factor and linkage analyses. Research Questions The study is designed to describe the dimensions used in five different groups by group members in formulating similarity judgments of other group members. Each group constitutes a separate replication. The three following research questions outline the aim of the data analysis. Question 1. How many dimensions best represent group members' similarity judgments of one another in each of the five groups sampled? Question 2. What are the identities of the dimensions derived from the similarity ratings in each group? 54 Question 3. Do group leaders and group participants differ in regard to the salience of particular dimensions in making similarity judgments? Analysis of the Data Similarity ratings from each group member are used to construct a similarity matrix with N(N—l)/2 ratio scale proximity judgments. The similarity matrix provided input for an individual differences multi- dimensional scaling analysis (INDSCAL) (Carroll and Chang, 1970). The analysis of the data involves three steps: (a) obtaining the INDSCAL output that would best represent the dimensions underlying the similar- ity ratings, (b) identification of the INDSCAL dimensions, and (c) com- parison of leader and participant subject weights on each dimension. The INDSCAL model postulates a common set of psychological dimen- sions that Carroll and Chang (1970) state "correspond to fundamental sensory, perceptual, or judgmental processes that vary in salience, or strength of effect on perception, across individuals" (p. 285). INDSCAL assumes a common set of factors underlying the perception of the stimuli. It also assumes the similarity judgments of each subject are linearly related to a weighted Euclidean distance. These assumptions are restric- tive but the model allows for verification of the assumptions. INDSCAL outputs two types of configurations. A Stimulus X Dimensions matrix defined the group stimulus space. The dimensions identify fixed reference axis that correspond to properties of the stimuli. The position of each stimuli as seen by all g! fi di ac thu Spa vei how con. coo: t01 fro: var: Car: orig his his full: the F Telev 55 group members is chartedcnleach axis. The number of dimensions that best fit members' ratings is determined by the additional amount of variance accounted for in one, two, three, or more dimensional solutions. The dimensions are ordered in correspondence to the relative variance accounted for. The axis are fixed and no rotation is allowed due to the judges weights, resulting in a direct interpretation of the particular axis derived by the analysis. A Subject X Dimensions matrix identifies an individual's subject space through subject weights. Each individual's perception of the stimuli is related to the group's perception by the weight matrix. The weights derived from the analysis are "stretching factors" that reveal how much a dimension is stretched or contracted relative to the group configuration to represent that particular subject's judgments. The coordinants of a given subject's point in the subject space correspond to the weights on the various dimension for that subject. The distance from the fixed origin corresponds directly to an approximation of the variance accounted for by that dimension for a particular subject. Carroll and Chang (1970) state that "one subject's being closer to the origin on that line would indicate simply that less of the variance in his data is accounted for by that common configuration (either because his data are noisier or because additional dimensions are needed to fully account for this subject's data)" (p. 297). INDSCAL describes the pattern of a subject's perceptual space that may relate to other relevant attributes. The reader is referred to Carroll and Chang (1970) or Subkoviak (1975) for a more technical treatment of INDSCAL. 56 The underlying dimensions that best represent subjects' similarity judgments are derived through INDSCAL. The identification of these derived dimensions is the second step of the analysis. Dimensions must be labeled through an external criteria since the similarity ratings asked for judgments based on the "most relevant" attributes perceived by the subjects. The GSD serves as an external criteria in an attempt to label the INDSCAL dimensions. Ratings on the GSD adjective pairs for each stimuli (group member) are obtained by averaging the n-ratings made by each member group. The GSD adjective pair ratings were then summed according to the original factor clusters (see Table 3.3). Spearman rank correlations were computed between the INDSCAL dimensions and the original factor groupings. A pg§£_hgg_analysis involved computing Spearman correlations between the INDSCAL dimensions and clusters de- fined through three alternative means of clustering the GSD adjective pairs. These three methods included the four and three cluster factor analysis solutions (see Table 3.3), and the clusters derived through the linkage analysis (see Table 3.4). The results of the analysis and pg§£_hgg_analysis are reported in Chapter IV. Through the derivation of a subject weight matrix, INDSCAL allows an analysis of individual differences. The INDSCAL model assumes that individuals differentially weigh the common dimensions underlying their perceptions. The final step of the data analysis is a comparison of leader and participant subject weights. The analysis of subject weights as a function of designated role is restricted by the small sample size; however, the subject weights for group leaders and group participants will be examined for similarities and dissimilarities. 57 Summary The design of the study is presented in this chapter. Detailed descriptions of the sample, measures, research questions, and analysis of data are presented. The aim of the study is to describe and identify dimensions assumed to underly group members perceptions of one another. Data were collected from a sample of five interpersonal process— oriented groups. Three of the groups were part of an undergraduate Psychology class and the remaining two groups were part of an under- graduate Education course. The groups associated with the Psychology course have been labeled groups A, B, and C. Groups D and E are the two groups from the Education course. Measures were administered during the eighth week of the groups' nine-week duration. Two types of measures were taken. Similarity ratings of all possible pairs of group members presented in a Ross (1934) ordering were completed first. Subjects were then asked to rate each member of their own group, including themselves, on sixteen adjective pairs of a Group Semantic Differential (GSD) instrument. The sixteen adjective pairs were composed of bipolar adjectives expected to describe four factors: dominance, affiliation, activity, and goal orientation. Two 222; Egg. analyses were performed on the GSD adjective pairs to check the validity of the proposed factor clusters. The first procedure was a factor analysis across all group members (N - 41) that solved for three and four cluster solutions. The factor analysis clusters had different 58 compositions from the original clusters. The second procedure was an elementary linkage analysis of the adjective pair ratings in each group that found differences between the groups in the formation of adjective pair clusters. The GSD clusters were designed to be an external criteria for labeling dimensions derived from an analysis of the simi— larity ratings. An individual differences multidimensional scaling analysis (INDSCAL) (Carroll and Chang, 1970) was used to analyze the similarity ratings made by each group member. INDSCAL derived fixed dimensions underlying the similarities data and charted the coordinates of each stimulus in a group subject space defined by those dimensions. Obtain- ing the number of dimensions that best represent the similarities data was the first purpose of the study. The identity of the dimensions was the second research question. The derived dimensions are correlated using Spearman's rank correlation with the GSD factor clusters in an attempt to label the INDSCAL dimension. INDSCAL also reports individual differences in the salience of each dimension. Differences between group leaders' and group participants' subject weights are explored and concern the studies final research question. The results of the data collection and analysis described in this chapter are the tapic of Chapter IV. CHAPTER IV ANALYSIS OF RESULTS The results of the present study are organized into three areas. The initial concern of the study was the dimensional representation of group members' interpersonal perceptions of one another. The dimen- sional representation was derived through an individual differences multidimensional scaling analysis (INDSCAL) (Carroll and Chang, 1970). The second area of results to be described involved the identification of the INDSCAL dimensions. Finally, the results of a comparison between leaders and participants' subject weights on the derived dimensions is described. Dimensional Representation Question 1. How many dimensions best represent the subjects' similarity judgments in each group? One, two, and three dimensional INDSCAL solutions to the similar- ity ratings were compared for each group. The amount of information recovered from the similarity ratings by one, two, and three dimensional INDSCAL solutions was indicated by the average subject correlation co- efficients computed between the INDSCAL scores and the original similar- ity data. These correlations are presented in Table 4.1. The higher 59 60 Table 4.1. Average Subject and Mean Square Correlation Coefficients Between Computed INDSCAL Scores and Original Similarity Data 3 dimensions .92 .84 .80 .64 .82 .67 .79 .63 .80 .64 2 dimensions .88 .78 .72 .55 .73 .55 .72 .54 .72 .53 1 dimension .82 .68 .61 .41 .62 .39 .57 .39 .53 .33 _. £1. {Wham-33.9.1 the correlation, the better the fit between the dimensional representa- tion of underlying structure and the input data. The correlations increase as the number of dimensions increase. The object is to obtain the best fit in as few dimensions as possible. The mean square corre- lation coefficient (r2) provides an estimate of the amount of the vari- ance in similarity ratings that is accounted for and is also presented in Table 4.1. A three-dimensional solution accounted for less than two-thirds of the total variance in similarity ratings in four of the groups. Similarity judgments involve a number of attributes in these groups. The remaining one-third of the total variance contains a number of dimensions accounting for a diminishing proportion of the variance. The four groups have a common pattern in the graduated amount of vari- ance accounted for by one-, two—, and three—dimensional representations of the data. The amount of total variance accounted for in these four groups averaged 38% for a one-dimensional solution, 542 for a two- 61 dimensional solution, and 65% for three dimensions. The amount of vari- ation from these averages was very slight across the four groups. The relative contribution of each of the three dimensions to the total vari- ance was more balanced than reported in previous research (Jackson et a1., 1957). The second and third dimension account, on the average, for 16% and 11% of the total variance respectively, however the major proportion of the variance remains attributed to the first dimension. The addi- tional information provided by the second and third dimension was large enough to judge the three-dimensional structure as best fitting the data. Consequently the three-dimensional structure was used for further anal- ysis. Group A was different than the other groups since the overall correspondence of the data in group A with the INDSCAL solution was much better than in the other groups. One dimension in group A accounted for over two-thirds of the total variance in the data. A second dimension increased the proportion of variance 102, and a third dimension increased the proportion 62. The members of group A were in much greater accord regarding the dimensions used for similarity judgments than were members of the other groups. The individual subject's judgments were very well represented by the group configuration defined by INDSCAL and suggests that subjects' perceptions were more cohesive than in the other groups. The three dimensional solution of group A was also used in further analysis. 62 Dimension Labels Question 2. What are the identities of the three INDSCAL dimen- sions derived from the similarity ratings in each group? Four different methods of clustering the sixteen GSD adjective pairs have been described in Chapter III. The various clusters will be referred to throughout this chapter as follows. The 3 priori four ? factor clusters are affiliation, dominance, activity, and goal orienta- « é; tion. The four factor—analysis clusters are designated affiliation', dominance', activity', and goal orientation'. The three factor-analysis clusters are referred to as affiliation", dominance/activity, and goal orientation". The linkage analysis clusters are designated by roman numerals I, II, or III followed by a letter indicating the group the cluster is from. The adjective pair clusters from the four methods of grouping were correlated with the INDSCAL dimension using Spearman's rank correlation coefficient. High negative correlations were as useful as high positive correlations in identifying the INDSCAL dimensions since the polarity of an INDSCAL dimension can be reversed in the results. The correlations between the sixteen adjective pairs and the INDSCAL dimensions are reported in Table 4.2. The GSD adjective pairs and adjective pair clus- ters were used to identify the INDSCAL dimensions. A discussion of the relationships between the GSD adjective pairs and clusters and the INDSCAL dimensions in each group follows. 63 we .. e .. . all“ . ‘Lh .oN. an.- ne.- o~.- n~.- so.- -.- am.- an.- do. no.- an.- as. so. no.1. sausages-sauna we. o~.- an.- no. o~.- as. as. we. nn.- s~.- so.- on.- No. as.- oo.- euueequ-auaoeuav-uu .n~. we.- on.- No.- oH.- as.- mm.- no.- mm.- 0 mm. on. e on. o~.- squeegee-aomoa Ho. mo.- no.- m~.- oo.- n«.- «a. «a. so.- as.- «o.- ea.- co. No. «a.- Haoeuouoa-xovoeuuo w-o No.- e~. ma. mN. we.- km.- Hm.- s~.- me. An. ea. no.- es. en.- asuueeuez-eo>ao>ea Wuo.- H~.- o~. no. we. we.- an.- no.- so.- He. em. as. ~e. oH.- No.- sauaem-osauasana mm~.- m~.- we. AN. ca. on.- ok.- en.- on.- me. an. n“. n~.- mN. n~.- scam-uaue jWo.- nn.- me. so. as. no.- an.- Ho.- q~.- on. on. No. H~.- eN. No. o>euuak-o>eou< nd~.- on.- so.- no. so.- as.- an.- nn.- as.- me. as.- em. as. an.- no.- eaten-oauaoo wko.- oo.- on. oo.- we. we.- an.- Ne.- Hq.- ms. «N. a». oa.- no. no.- haeeoeuues-maeeoeum M ed. ak.- No. no.- o~.- os.- an.- ms.- on.- an. o as. oo.- e~.- -.- eaoo-sunz -m~.- Nm.- am. an.- ms.- NM.- am.- -.- n~.- an. -.- on. o~.- Ne. an.- unauuofiua-uaaoeouu< M-m~.- an. mm. mm.- an. -.- -.- so. on. an.- es. as.- he.- as. as. concedes-nouns: l-mm.- He.- an. mo. em. Ne.- am.- Hw.- s~.- no. so. no. “,mn.- mm. MN. usoaaom-uemua -a~.- cs.- oa. mo. we. o~.- mn.- He.- o~.- No.- es. on. i so.- NN. no. oaoeuuaon-uaoeauaoe=H +,eo.- Na.- an. Ho. he. so.- he.- sm.- os.- so.- sh. an. MMN.- as. as. sues-mcotum .ir .Mr - mn No no ”no No so me No He no No no no no on _ m -w a o m < manomo mzoemzmxao geomoze moHoomnsm emu soouxdm one one dsouu zoom a“ msoamooaan Aanswer questions regarding what types of groups choose particular dimensions of perception. The characteristics of groups using particular dimensions are central to an attempt to deter- mine why particular dimensions are chosen. For example, group A in the present study stood out from other groups studied due to the salience of a single dimension. The group attributes corresponding with this particular pattern of dimensional representation may have been examined by a more comprehensive monitoring of the group behavior. The development of the dimensions of interpersonal perception are another area of concern. Future research designs should collect data at regular intervals during the group life instead of just at one time. Changes in the use and importance of various dimensions over time can 96 yield information relevant to interpersonal behavior in the group. A comprehensive study must monitor the relation of changes in group process with changes in interpersonal perception. A critical incidents method can be used to assess group development and such data may berelat- ed to the dimensions of perception being used by a group at a particular point in time. Through these means, many of the assumptions about group members' perceptions in process-oriented groups during different develop- mental stages may be tested. The importance of individuals who are seen as similar or dis- similar in the group's perceptual field can be related to a variety of sociometric choices. Measurements of variables such as amount of self- disclosure, selection of friends, selection of working partners, and amount of interaction may be predicted by the similarity or dissimilar- ity of group members on dimensions in the common group space. Jones and Young (1972) reported accurate predictions from the group perceptual field computed by INDSCAL on the selection of committee members and on people socialized with. The additional power derived from using dimen- sions relevant to the group members appears to have predictive effici- ency. A larger sample of groups will also allow an assessment of the effects of leadership styles upon the dimensions group members use to represent one another. The effects of directive versus nondirective leadership styles upon the dimensions of interpersonal perception chosen by the group would be of interest as an assessment of the influence of leadership style in groups. 97 The GSD used in the present study needs further validation and reliability studies. A broader range of adjective pairs should be examined to establish reliable subscales for the various factors that are being scaled. An elaboration of this measure will provide a more precise external criteria for use in describing the INDSCAL dimensions. An equal number of subscales should be used to evaluate each factor cluster. The range of dimensions of perception used in process-oriented groups has not been identified. The GSD should be supplemented with other measures assessing attributes of the stimulus persons in order to attain a good identification of the INDSCAL dimensions derived from group members similarity ratings. The Likert scales used on the GSD should be modified in future research that performs rank correlations. Many subjects restricted their use of the Likert scale to three or four intervals in spite of directions asking them to "use the entire scale." A procedure requesting that subjects rank order all stimuli on each bipolar adjective scale would be one means of assuring full variability in the ratings. The precision of the GSD instrument would subsequently be increased. The group stimulus space derived from INDSCAL has descriptive and predictive value. The most striking finding of the present study was that subjects in different groups use different dimensions in making similarity ratings. Further research regarding to what these group differences are related is needed. The group stimulus space may also be used to predict outcome and group process variables. The fact that 98 the dimensions are relevant to the group members perceptions may in— crease the power of this procedure in providing an effective predictor of group behavior and outcomes. BIBLIOGRAPHY BIBLIOGRAPHY Adams, H. B. "Mental Illness" or interpersonal behavior? American Psychologist, 1964, 19: 191-97. Bales, R. F. Interaction Process Analysis. Cambridge, Mass.: Addison Wesley, 1950. . Personality and Interpersonal Behavior. New York: Holt Rinehart, 1970. Bieri, J., Atkins, A. L., et al. Clinical and Social Judgments: The Discrimination of Behavioral Information. Huntington, New York: Krieger Publishing Co., 1975. Bryson, J. B. Factor analysis of impression formation process. Journal of Personality and Social Psychology, 1974, 30: 134-43. Burke, R. L. and Bennis, W. G. Changes in perception of self and others during human relations training. Human Relations, 1961, 14: 165-182. Carson, R. C. Interaction Concepts of Personality. Chicago: Aldine Publishing, 1969. Carroll, J. D. Individual differences and multidimensional scaling. In Multidimensional Scaling, Vol. 1, Shepard, R. N., Romeny, A. K., and Nerlove, S. B. (eds.). New York: Seminar Press, 1972. Carroll, J. D., and Chang, J. Analysis of individual differences in multidimensional scaling via an N-way generalization of Eckart- Young decomposition. Psychometrika, 1970, 35: 283-319. Carroll, J. D., and Chang, J. J. How to use INDSCAL, a computer pro- gram for canonical decomposition of N-way tables and individual differences in multidimensional scaling. Murray Hill, N. J.: Bell Laboratories, 1972. Cohen, A. M. and Smith, R. D. The critical incident in growth groups: Theory and technique. LaJolla, Calif.: University Associates, 1976. Cohen, A. M. and Smith, R. D. The critical incident in growth groups: A manual for groupgleaders. LaJolla, Calif.: University Associates, 1976. 99 100 Cronbach, L. J. Proposals leading to analytic treatment of social per- ception scores. InIL.Tagiuri and L. Petrullo (Eds.), Person per- ception and interpersonal behavior. Stanford, Calif.: Stanford University Press, 1958. Fiedler, F. Interpersonal perception and group effectiveness. In R. Taguiri and L. Petrullo (Eds.), Person perception and inter- personal behavior. Stanford, Calif.: Stanford University Press, 1958. Finn, R. H. Effects of some variations in rating scale characteristics on the mean and reliability of ratings. Educational and Psycho- logical Measurement, 1972, 32: 255-65. Foa, U. G. Convergences in the analysis of the structure of inter- personal behavior. PsycholpgyfReview, 1961, 68: 341-53. Green, P. E. and Rao, U. R. Rating scales and information recovery-- How many scales and response categories to use? Journal of Marketing, 1970, 34(3): 33-39. Hartley, J. A. A semantic differential for assessing group process changes. Journal of Clinical Psychology, 1968, 24: 74. Hastorf, A. H., Richardson, S. A., and Dornbusch, S. M. The problem of relevance in the study of person perception. In R. T. Taguiri and L. Petrullo (Eds.), Person Perception and Interpersonal Behavior. Stanford, Calif.: Stanford University Press, 1958: 54-62. Hurley, J. R. Helpful behaviors in groups of mental health profession- als and undergraduates. International Journal of Group Psycho- therapy, 1976, 26: 173-89. Hurley, J. R. Only two major interpersonal dimensions relevant to psychotherapy? Paper presented at the 33rd Annual Conference of the American Group Psychotherapy Association, Boston, 1975. Hurley, J. R. and Force, E. J. T-group gains in acceptance of self and others. International Journal of Group Psychotherapy, 1973, 23: 166-76. Jackson, D. N. The multidimensional nature of human social perception. Canadian Journal of Behavioral Science, 1969, 1: 227-62. Jackson, D. N., Messick, S. J., and Solley, C. M. A multidimensional approach to the perception of personality. The Journal of Psychology, 1957, 44: 311-318. 101 Jones, L. E. and Young, F. w. Structure of a social environment: longitudinal individual differences scaling of an intact group. Journal of Personality‘and Social Psychology, 1972, 24: 108-121. Kelly, G. A. The Psychology of Personal Constructs. New York: Norton, 1955. Leary, T. Interpersonal Diagnosis of Personality. New York: Ronald Press, 1957. Lewis, P., Lissitz, R. N., and Jones, C. L. Assessment of change in interpersonal perception in a T-group using individual differences multidimensional scaling. Journal of Counseling Psychology, 1975, Lopis, J. E. Group process and interpersonal communication strategies for teachers. Michigan State University: Silver Burdett Co., 1975. MacLeod, R. B. The phenomenological approach to social psychology. In R. Tagiuri and L. Petrullo (Eds.), Person Perception and Inter- personal Behavior. Stanford, Calif.: Stanford University Press, 1958. Maguire, T. 0. Semantic differential methodology for the structuring of attitudes. American Educational Research Journal, 1973, 10: 295-306. Matell, M. S. and Jacoby, J. Is there an optimal number of alternatives for Likert scale items? Effects of t-sting time and scale proper- ties. Journal of Applied Psychology, 1972, 56: 506-09. McQuitty, L. L. Elementary linkage analysis for isolating orthogonal and oblique types and typal relevancies. Educational and Psycho- logical Measurement, 1957, 17: 207-229. Miller, G. The magic number 7 :12: some limits on our capacity for processing information. The Psychological Review, 1956, 63: 81- 97. Murray, H. A., gt El- Explorations in Personality. New York: Oxford University Press, 1938. Osgood, C. E., Suci, G. J., and Tannenbaum, P. H. The Measurement of Meanipg. Urbana, 111.: University of Illinois Press, 1957. Ross, R. T. Optimum orders for the presentation of pairs in the method of paired comparisons. Journal of Educational Psychology, 1934, 25: 375-382. Subkoviak, M. J. The use of multidimensional scaling in educational research. Review of Educational Research, 1975, 45, 3: 387-423. 102 Tagiuri, R. and Petrullo, L. (Eds.). Person Perception and Interper- sonal Behavior. Stanford, Calif.: Stanford University Press, 1958. Thompson, E. C. An idiographic analysis of trait integration processes in impression formation. Unpublished doctoral thesis, Michigan State University, 1976. Tuckman, B. w. Developmental sequence in small groups. Psychological Bulletin, 1965, 63: 384-99. APPENDIX A APPENDIX A Instructions: We are interested in the views that you have concerning the members of this group. Please rate the similarity of the group members according to what you feel are the most relevant attributes of the individuals by checking the number that best represents your feeling about each of the following pairs of group members. MW. mm 1 6 2 * ____ .___ ____ ____ ____ .___ 1 2 3 4 5 6 7 9 6 3 ___. ___. ___ .... l 2 3 4 5 6 7 8 6 4 ___ .___ ____ ____ ____ ___. ___ l 2 3 4 5 6 7 765 __ __ __ __ _— 1 2 3 4 5 6 7 6 a] __u .__ ___ . ___ ___ 1 2 3 4 S 6 7 3 6 2 ____ ____ ____ ____ ___. ____ .___ 2 3 4 5 6 7 4 6 9 ____ ____ ‘___ .___ ____ ____ .___ 1 2 3 4 5 6 7 5 6 8 ____ ___- .___ ___. l 2 3 4 5 6 7 6 a7 __, ___ ___ ... ___ ... 1 2 3 4 5 6 7 1 a3 ___ ___ ___ __. __. 1 3 4 5 6 7 2 6 4 ____ ____ ____ ____ ____ ___ ____ 1 2 3 4 S 6 7 9 6 5 ____ ____ ____ -___ ____ .___ ____ 1 2 3 4 S 6 7 8 6 6 ____ _., .___ ___. ___. 1 2 3 4 5 6 7 7 6 1 ___ .___ ____ ___ ___ ___ ___ l 2 3 4 5 6 7 4 6 3 _“_ ____ .___ ___ .___ ____ ,___ 1 2 4 5 * Subject pairs 103 \ egg mssnum 1 2 T T T T T T T T T 'T T T T T T ___. T T T T T T T T T T T T T T T T T T T T T T 104 “l wl wl a! “I bl TI 5| “I §| Tl bl Tl bl bl mi M' ml ml ml ml ml ml “i ml ml o' ol OI GI NI Nl ~| “I ~l Vl “I N. “I NI 105 Describe each group member, including yourself, on the following scales. Be as discriminating as possible between individuals by making use of the entire scale. 31mg: :11: AK 1 * ‘1" T‘ T T T T T 2 2 T T T T T 3 1 2 3 T T T "'7'" 4 2 T r. 5 T T 5 2 T t. 5 Tc— 7 6 1 ‘2” T ’T T T T 7 1 T T T T T T 8 1 T T T T T T 9 T T T T T T T T T T T T T T T T T T T ‘6" 'T‘ * name of group menber PASS IVE 106 mi °~i ACTIVE 107 ACCEPTING REJECTING T T T T T T T T T T T T T T T T T T T T T T T T T T T T ’T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T 108 FOLLOWS LEADS T T T3T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T ‘6" T T T T T T T T 109 DISLOYAL 110 HARSR 111 2223?.13 2% H ml Ln) U! 0‘ N .... N La) 1.x u: c N N w b Ln 0\ N 112 FAQT SLOW H I» U b VI 0‘ \l TRADITIONAL T T T T T T T T T T T T T T T T T T T T T T 113 RADICAL INDEPENDENT 114 DEPENDENT 115 WARM 116 mm m N U ;> u C‘ \J L» b1 0‘ 117 TALKAIIVE SILENT ~-— -—~ -_ —-—- ———-——. —-_ m l HERETI CAL 118 ORTHODOX UNF‘RIEEIDLY 119 FRIENDLY 120 INVOLVED mmuwu 1 __ __ __ __ _ ___. __ 1 2 3 z. 6 7 2 __ __ __ ___. __ __ __ 1 2 3 4 5 6 7 3 __ __ __ __ ___. __ __ 1 2 3 4 5 6 7 z. __ __ __ __ __ __ __ 1 2 3 I. 5 6 7 5 __ __ __ __ __ __ __ 1 2 3 4 s 6 7 6 __ __ __ __ __ __ __ 1 2 3 4 s 6 7 7 __ __ __ __ __ __ __ 1 2 3 4 s 6 7 T T T T T T T T T T T T T ‘7" T T T T T T T T T T T T T T