AR APPLICATEON OF THE MULTiPLE ATTRéBUTE MEASUREMENT MODEL: MEASSREMENT MB MAREPULATEGN CF SOURCE GRED§B§L§TY “Ti esis for the Degree 0f H A. A xiv-Eb!“ 31W EVE 3;".[7‘133 ML h 42‘... . . 69:“! A 3.375 ' H '23 {\— :9 3 111(1 U3“ L) r# I I { L11: Univcxsit W’T‘ I,IBRARY hmm State Lt" _—‘ “.17Rl'J'I‘ ' 1 O \ ";..1- _ represe of each tnat di are fun. the Hm of a do! Tn ABSTRACT AN APPLICATION OF THE MULTIPLE ATTRIBUTE MEASUREMENT MODEL: MEASUREMENT AND MANIPULATION OF SOURCE CREDIBILITY By Michael J. Cody A dimensional model of judgment was proposed wherein concepts are represented as points in a multidimensional space, and the projections of each concept on the dimensions represent the values or magnitudes on that dimension. Attributes are held to be distinct from dimensions, and are fundamentally scaled into a multidimensional space comprising both the linguistic descriptors of a domain (trait adjectives) and the objects of a domain (person-concepts). The model was applied to the measurement of the source credibility construct by scaling into a multidimensional space trait adjectives (attributes) used typically as the bipolars of unidimensional scales, public figures and an ideal point (i.e., "Ideal Credible Source"). The model provides the following types of information: (l) A score for the ideal point on each attribute; (2) the salience of each attribute for the "Ideal Credible Source"; (3) a score for each public figure on each attribute; and, (4) the saliency of the attribute in perceiving each pub- lic figure. The thesis presented results concerning the following char- acteristics of the model: (l) the observed stability of the location of each of the "stable concepts" (the linguistic descriptors of the domain) in the multidimensional space was substantially high; (2) a considerable number of dimensions were found to be reliable; and (3) the attribute sa- lience measure was supported. will each lic i Credi occur the cc 1 contra speech cedures moved t "Egativu contrar; eXpiaim with the ARUIGtE( ti0n of additior. A an jr Based on the simple assumption that associatively linked concepts will converge (or, that disassociatively linked concepts will "repell" each other), messages were designed to move a relatively unfamiliar pub- ,lic figure closer to the "Ideal Credible Source" (us away from the "Ideal Credible Source." Further, it was predicted that these motions would occur along specific vectors -- the resultant of the vectors representing the concepts used in the message. To test these hypotheses, four posttest groups were used: (l) a control group; (2) a Speech only group (who received only a persuasive speech attributed to the manipulated public figure and received no cred- ibility induction message); (3) a positive induction message group; and, (4) a negative induction message group. Results indicated that the pro- cedures employed in the positive induction message group successfully moved the public figure along the predicted vector. The results of the negative induction message group indicated that the public figure moved contrary to the hypothesized vector. Failure to predict this motion was explained in terms of confounding the effects of the induction message with the effects of the speech, and by the fact that the unfamiliar (man- ipulated) public figure's motion was partially interrelated with the mo- tion of a public figure to whom he was linked. To obtain a manipulation check on the induction of higher and lower levels of credibility, a topic was selected from an attitude pretest and a persuasive speech was attributed to the manipulated public figure. Results of the manipulation check indicated that the persuasive speech was persuasive by itself and that the credibility inductions had little additional impact on the amount of attitude change observed. While there is an indication that the amount of attitude change obtained in the negat only 4 signii negative induction message group was less than that in either the speech only and positive induction message group, these differences were not significant. AN APPLICATION OF THE MULTIPLE ATTRIBUTE MEASUREMENT MODEL: MEASUREMENT AND MANIPULATION OF SOURCE CREDIBILITY By ‘\ Michael JIvCody A THESIS Submitted to Michigan State University in partial fulfillment for the degree of MASTER OF ARTS Department of Communication 1976 Accepted by the faculty of the Department of Communication, College of Communication Arts, Michigan State University, in partial fulfillment of the require- ments fOr the Master of Arts degree. ,4 c ., -» f ' - .' -’. .o- Chai man, Ll. 411011 3 ”I. (I .12". (A [I --((’f 4.4.13 7/ HM I Q I nus-g. haus, valuab Katrini mental bleak.y ACKNOWLEDGMENTS Recognition goes first to the author's advisor, Dr. Erwin Betting- haus, whose time, patience and perceptive criticism were extremely valuable in the preparation of the thesis. Recognition also goes to Dr. Katrina Simmons for her useful comments on the text, comments on experi- mental procedures used, and for her encouragement when times were rather . bleak. Special acknowledgment must be given to Dr. Joseph Noelfel. With- out his help the thesis could never have been attempted. His help, I feel, should be accurately recognized and thanked: (l) he started the work on metric multidimensional scaling which made the thesis possible; (2) he helped on the design of the present study; (3) he wrote the equa- tions for rotation, message design and message effectiveness; and (4) he he spent many hours with the author testing (by hand) equations. John Marlier also deserves special acknowledgment. The papers we co-authored lead to the present thesis. Indeed, parts of the thesis have leaned heavily on our earlier work. I thank him for the help on writing, particularly writing the messages used in the study. I thank him also for helping in data collection during the posttest phase of the study. Without John Marlier's help and encouragement, the goals of the present study would have been fewer and certainly less general. Thanks go to Ken Nyhagian for help in collecting data and coding data, and to Kathy Sherry for helping to collect data, for keypunching and for typing. Thanks should be given to the following people for telephoning $5 for the posttest phase of the study (or for coding): Gary ii |_ I l- Noble Marill Heiser like t Finall Cooney blance while C: Arts p sis. Connun' that ii data an A I” WA "3 Chapter 'A RA in fini for eac fatter tlon pr (4) I t“ him f0r thank D‘ h: "m for COrn91a‘ / 1/4 TC) f_ (A’Eég \:) ,. /C.‘ ‘Gzz)><~CDC? ‘L Noble, Elena Yang, Jan Schesinger, Don Hayes, Alan Muskovitz, Pati Kluck, Marilyn Miles, Susan Zimmerman, Terri Wasylenko, Genese Piggott, Marie Meisenback, Kathryn Ulmer, Terry Vujea and Delann Williams. I would also like to thank Marilyn Williams for an excellent job in keypunching. Finally, I want to thank several close friends, Teri Albrecht, Jackie Ur- Cooney and Carter Clary, fbr encouragement necessary to maintain a sem- blance of sanity (on my part) while the study was being conducted and while data were being analyzed. Communication Research Services and the College of Communication Arts provided services that were critical for finishing the data analy- sis. I thank the College for hundreds of dollars of computer money, and Communication Research Services for the additional hundreds of dollars that it provided. Roughly one thousand dollars went into completing the data analysis and these services are gratefully acknowledged. Kathy Clyde deserves recognition for the graphics in the thesis and for typing the thesis. I thank her for being so cheerful when all of Chapter III had to be retyped. This cheerful attitude made a difference. Richard Holmes must be identified and thanked for his crucial role in finishing "our thesis." I would like to acknowledge and thank him for each specific task he performed: (1) I thank him for programming factor correlations; (2) I thank him for writing the cross-group correla- tion program; (3) I thank him for writing the "Marlier's R" program; (4) I thank him for writing the attribute salience program; (5) I thank him for rewriting the "Rotate" subroutine in the Galileo program; (6) I thank him for writing the “message generator" program; and (7) I thank him fOr writing the "message effectiveness“ program (with motion vector correlations). I further thank Rick for running test decks through these iii _ r-\ j r‘. I , K4. D - L l - PIOQT and f quest the p' the fi programs, for helping to test equations (by hand) far into the night,, and for being a drinking partner. However, in compliance with Rick's re- quest on how my acknowledgment to him should be phrased, please disregard the preceding acknowledgment and note that all I really want to say is the following: "Rick helped too." iv I‘ll II III IV Chapter I II III IV TABLE OF CONTENTS INTRODUCTION: THE PROBLEM OF MEASURING SOURCE CREDIBILITY ........................................... Assumptions of the Semantic Differential ................. Assumptions of the Multiple Attribute Measurement Model.. On the Assumptions of the Proposed Model ................. (a) Definition of "Implicit Personality Theory" ....... Eb) Generality of "Implicit Personality Theory" ....... c The Nature of the Phenomena ...................... Derivation of Multiple Attribute Model ................... Summary and Hypotheses ................................... METHODS .................................................. Subjects ................................................. Sorting Task and Criterion Pair Selection ................ Selection of Public Figures .............................. Topic Selection Pretest .................................. Message Design Pretest ................................... Posttest ................................................. a) Materials ......................................... b) Procedures ........................................ (c) Randomization Check ............................... RESULTS .................................................. Comparability of the Spaces .............................. (a) Procedures Utilized to Adjust the Data ............ (b Overall Measure of Fit ............................ c Cross-Group Correlations .......................... (d) Factor Correlations ............................... Attribute Salience ....................................... Hypotheses ............................................... Manipulation Check ....................................... Message Effectiveness .................................... DISCUSSION ............................................... Overview of the Measurement Model ........................ Overview of the Effects of the Experimental Manipulations ......................................... Directions for Future Research ........................... Page 41 41 43 44 48 55 58 58 6O 60 117 117 l2] 127 BIBLI APPEN TABLE OF CONTENTS (cont'd) Page BIBLIOGRAPHY ..................................................... 130 APPENDIX ......................................................... 138 vi Tabie Table 10 II 12 l3 I4 15 16 I7 18 19 20 LIST OF TABLES 1333; Demographic Information on Ss Utilized in Pretest on Message Design and Four Posttest Groups ................... 42 Results of Sorting Task ................................... 45-47 Results (Familiarity and Importance) of Public Figure Selection Pretest ......................................... 49 Results (Informational Sources) of Public Figure Selection Pretest ......................................... 50 Results (Identification of Occupations) of Public Figure Selection Pretest .................................. 51-52 Results of Topic Selection Pretest ........................ 53-54 Results of Means Trimming Procedure on the Birch Bayh- Ideal Credible Source Distance Distribution .............. 65 Correlations and ;_Scores for Stable Concepts ............. 67 Cross-Group Correlations between Each Stable Concept ...... 70 Results of Factor Correlation Procedure ................... 74 Salience Scores for Birch Bayh ............................ 77 Salience Scores for Hubert Humphrey ....................... 78 Salience Scores for Ideal Credible Source ................. 79 Salience Scores for George McGovern ....................... 80 Correlations between Salience Scores and Ratings on Relevance and Importance Scales ........................... 82-83 Results of One-Way Analysis of Variance ................... 85 Analysis of Variance for Attitude Change Scores ........... 87 Control Group Coordinates ................................. 90-91 Speech Only Group Coordinates ............................. 92-93 Positive Induction Group Coordinates ...................... 94-95 vii Tab' 21 22 23 24 25 26 LIST OF TABLES (cont'd) Table Page 21 Negative Induction Group Coordinates ........................ 96 22 Correlations and Angles for Bayh's Motion and the Resultant Vector ........................................ lO4 23 Correlations and Angles for Humphrey's Motion and the Six Negative Trait Descriptors .......................... l07 24 Correlations Between Humphrey and Bayh's Motion Vectors ..... lll 25 Correlations and Angles between the Bayh Motion Vector and the Resultant Vector in the Speech Only-Positive Induction Message Comparison and the Speech Only- Negative Induction Message Comparison ....................... ll4 26 Correlations and Angles between Humphrey and Bayh's Motion Vectors (Speech Only-Positive Induction Group Comparison and Speech Only-Negative Induction Group Comparison) ................................................. l16 viii F LIST OF FIGURES Figure Page 1 Hypothetical Illustration of Multiple Attribute "Scale"... 3l 2 .......................................................... 32 3 .......................................................... 33 4 .......................................................... 35 5 Illustration of Four-Pair Vector Solution ................. 57 6 Three-Dimensional Representation of Control Group- Speech Only Group Comparison .............................. 98 7 Three-Dimensional Representation of Control Group- Positive Induction Group Comparison ....................... 99 8 Three-Dimensional Representation of Control Group- Negative Induction Group Comparison ....................... 100 ix PPESEI latini e ical I analys Into I Partic One. t tained Struct the n‘ HCLaUg "Ideal CHAPTER I INTRODUCTION: THE PROBLEM OF MEASURING "SOURCE CREDIBILITY" Many of the constructs in the nomological networks that comprise communication theories are social perception constructs. Such con- structs as source credibility (Berlo, Lemert and Mertz, l969), source valence (McCroskey, Jensen and Valencia, 1973) and homophily (Rogers and Shoemaker, l97l) are key constructs in our theory building. The present study is primarily concerned with the general problem of re- lating these abstract constructs to measurement in the social and phys- ical reality. Such constructs are generally thought to be multidimensional by most recent investigators. Typically, to assess an object that is mul- tidimensional, covariances among ratings on multiple unidimensional scales are factor analyzed and factor indices are constructed from these analyses. Such procedures have led to advances in scientific inquiry into the nature and effects of complex and multidimensional constructs, particularly after the initiation of high speed computers. Several problems, however, continue to plague this approach. For one, the scales selected by the §_determine the factors that will be ob- tained, and this does not insure relevance of the factors to the con- struct. Further, as McLaughlin (1975) argued, it is not known whether the n-number of factors obtained are exhaustive of relevant factors. McLaughlin recommended locating a concept such as "Most Believable" or "Ideal Credible Source" into a multidimensional space of public figures. I [O 5); Credi point does Credibility would then be a simple function of distance from this ideal point. This method would provide a good measure of credibility, but does not answer the traditional theoretical question as to which factors of perceiving others are critical credibility factors. Utilizing a more traditional approach, McCroskey, Jensen and Todd (1973) attempted to answer the question by using factor scores to pre- dict to Likert-type items that purport to tap "Communication-Related Behaviors." They obtained multiple correlations of only .5 to .7, which implies either that the criterion variables did not differentiate credi- bility, that some relevant dimension(s) may not have been tapped, or perhaps that a credible source is one who does ggt_score consistently high on all factors. Heston (1973) demonstrated the viability of the argument that the "Ideal Credible Source“ may not be the source that loads consistently high. She reported the surprising results that the ideal source "...would be highly responsible, reliable, honest, just, kind, cooperative, nice, pleasant, sociable, cheerfu1, friendly, good- natured, and relaxed, and ggly slightly expert, virtuous, refined, calm, composed, verbal, mild, extroverted, bold and talkative" (p. 10, emphasis mine). Taken together, these considerations lead to the conclusion that an alternative measurement model for the source credibility construct should be developed. McLaughlin's model (1975) provides a global score which purports to reflect a multiplicity of receiver attributions to sources, but which provides no ready means of identifying which attribu- tions are made or the relative saliency of those attributions. The fac- tor analytic models allow identification of attributes salient to per- ceptions of credibility, but are not readily amalgamated into a mean ties that dica bute and s veloc elimi this facto. ASSursJ commuv USE av Indee: and ti equal. the d: Scale neCESs deDCe that . by de< dEVeh nature meaningful index of credibility as one complex construct. The difficul- ties in indexing scores on credibility factors stem both from the fact that middle-range scores on some credibility-related attributes may in- dicate maximal credibility, and from limitations imposed on the attri- bute configuration in a factor space by assumptions of factor analysis and semantic differentiation. It is the purpose of this paper to de- velop and test an alternative model which can combine the strengths and eliminate the weaknesses of the two models discussed above. Toward this end, it will be useful to examine the assumptions upon which the factor analytic model rests. Assumptions of the Semantic Differential While the Semantic Differential has been extensively employed in communication research, several of the key assumptions underlying its use are questionable. First, the approach assumes that the scale is indeed unidimensional, there exists some (center) point of neutrality and that the distance between each of the end points and the center are equal. Further, the lengths of each attribute scale are standardized; the distances between all pairs of bipolar adjectives are equal, and scale intervals are assumed to be equal categories. Finally, it is necessarily implied that any scale attempts to achieve some correspon- dence, some isomorphism, between the numbering system in the scale with that inherent or latent in the psychological continuum. One of the central limitations of unidimensional scales is that, by definition, they measure only one attribute, and factor analysis was developed precisely because objects of cognition are multidimensional in nature. Consider the typical factor analysis experiment: the §_selects a se ment struo ates relat the p proce repre a set of attribute scales, presumably exhaustive of dimensions of judg- ment in a particular domain, or to investigate some theoretical con- struct. Ss evaluate a number of concepts on these scales, which gener- ates a matrix of scores and ultimately a correlation matrix. The cor- relation matrix is factor analyzed by any usual procedure to determine the projections of the stimuli on r orthogonal axes. The goal of the procedure is to present a parsimonious representation of the data to represent the factors, or dimensions, of judgment. The development of factor theory was dependent upon assumptions of a common origin, bipolarity, equidistance of scale anchors from origin, and standardization of scale metric. The assumption of a common origin implies that (l) the centroid from which vectors originate is a point of neutrality, and (2) all vectors originate from this meaningful neutral location; that is, all attributes intersect at a meaningful lo- cation. The strong version of the assumption holds that these facets are necessarily true. Osgood gt 91, (1957) made this explicit and argued that intensity and direction are indicated by factor loadings. The weaker version of this assumption is never fully discussed in factor analytic research. It is made possible by arguing that the centroid is not necessarily a meaningful point of neutrality; that the sole reason that vectors originate at the origin is mathematical convenience. Hence, the weaker assumption of common origin only asserts that all attribute-line segments in the space intersect at the origin (or at some point, in the case of transformation), and not that there is any Special significance to the centroid. The "meaningful origin“ aSpect of the assumption of common origin is directly related to the assumptions of bipolarity and equidistance from the origin: One of the difficult methodological problems we have faced - unsuccessfully so far - is to demonstrate that the polar terms we now use are true psychologic- al opposites; i.e., fall at equal distances from the origin of the semantic space and in opposite direc- tions along a single straight line through the origin. And why use the adjectives? We assume that it is the lexical (root) meanings of our polar terms that determine judgments; adjectives are merely the most eneral and natural qualifiers in English. Osgood gt_§1,, 1957, pp. 327-328) Several studies have focused on this "difficult methodological" problem of bipolarity and equidistance. Wishner (1960) argued that one of the bipolar adjectives may be the grammatical opposite of the other, yet possess positive or negative implications of its own. This implies that the meaning of an adjective is not necessarily strictly defined as the opposite of its grammatical antonym, but by its set of formal rela- tions of implicating similarities and dissimilarities, with all other traits and concepts. More stringent tests of both bipolarity and equidistance assumptions have been offered by multidimensional scaling analyses. In testing the assumption of bipolarity, Anderson (1970) and Danes and Woelfel (1975) argued that line segments drawn from the centroid to each of two bipolars should have an angle between them equal to 180°. The fact that neither study found angles of 180° between these line segments supports Wishner's contention that each trait adjective possesses its own unique set of formal relations with other traits since, in M05, the location of a trait is dependent upon its perceived similarities with all traits. The gram- matical opposite is only one of many traits used as a reference point in the location of a trait. Both Anderson (1970) and Danes and Woelfel (1975) also assessed the common origin and equidistance assumptions by computing the distance bet the: taii Thes tion to d comnl space and ( vary dissii and at atelyy all t.‘ Commor arblti 0f the 0f the Rathe, assung I ”935‘“ VaIS, Values that F 0Des f between each concept point from the origin. If equidistance exists, then a ratio of smaller distance to the larger should be 1.00. The ob- tained ratios, in both studies, failed to support this assumption. These studies, then, would seem to indicate that the theoretic assump- tion of equidistance of bipolars from a common origin does not conform to data collected to test it. Hence, the invalidity of the strong version of the assumption of a common origin is demonstrated by: (l) locating individual points in the space vs. locating pairs of grammatical opposites in the space jointly; and (2) allowing the distances from each pair of grammatical antonyms to vary in length as a free parameter according to Ss' perceptions of dissimilarities vs. constraining all attribute line segments to equal and arbitrary length. It can be concluded that meaning is more accur- ately conceptualized as the result of the sum of compound reactions to all traits taken singularly and, secondly, that standard length and common differentiation of semantic differential scales impose severe and arbitrary constraints on measuring the meaning of a concept. In light of these conclusions, adoption of the “weaker version" of the assumption of the common origin for mathematical convenience must also be rejected. Rather a representation of semantic space which makes no assumptions, or assumptions more commensurate with available data, should be sought. Finally, the interval quality of the scale can be questioned. Messick (1957) found high correlations between obtained and assumed inter- vals, but quickly pointed out that due to restrictions on variation of values, such a relationship by nature must be very high. Messick found that positive intervals were consistently larger than symmetric negative ones for all scales. Interval distortions may not only be based solely on wh be in 127 IS! grOT reas impl with Chang ues ( stint point icall attri proje anal) Weani faCtL Stanc a Ira ing n SlnCe Peldt on the effects of positive ratings; there may also be “end effects“ -- where the tendency for extreme categories both positive and negative to be larger than the center ones. Gulliksen (1958) pointed out that on many of the individual items in the Measurement of Meaning (Osgood, Suci and Tannenbaum, 1957, p. 127) the variance approached zero. Gulliksen asserted: ”Clearly, it is not possible to determine accuracy of measurement when such a coarse grouping is used. For any measurement one needs a unit so fine that a reasonable determinatknnof error is possible" (p. 116). The two relevant implications are that a more precise scaling device is needed and that without accurate measurement there can be no accurate measurement of change. Additionally, it may be noted that low variances in scaled val- ues of stimuli may result also from "ceiling effects" resulting from stimuli being perceived by Ss as having projections beyond the end point of the presented attribute scale. Factor analysis cannot empir- ically test this possibility because it constrains the arrangement of attributes such that a stimulus which projects on one attribute must project onto all attributes. In conclusion, the validity of the assumptions upon which factor analysis of unidimensional scales rests is questionable. First, both a meaningful origin and the equidistance from origin condition are arti- factual, stemming from the forced association of pairs of points and standard lengths between end point. Secondly, assuming the meaning of a trait to be the opposite of its grammatical antonym and conceptualiz- ing meaning as a compound reaction to bipolar terms is questionable since the meaning of each individual trait is uniquely defined by its relation with all other traits. Linguistic determinancy of a trait's loca‘ in t: Assur cepts stabl cedur sampl singl throu dimeni exists LIC ur [\i mlned tial r ‘- LIC Uh USage, QUIStj h'"SIui haVlng location in the space can only be realized as the result of its location in terms of its formal relations with all traits. Assumptions of the Multiple Attribute Measurement Model The alternative representation of "semantic Space" may be described as a multi-dimensional array of linguistic elements (descriptor con- cepts, including unidimensional scale anchors). This configuration is stable in a Space generated through metric multidimensional scaling pro- cedures from aggregated data of a sample of language users (i.e., a sample of §§ who share a common language). Such an array constitutes a Single multidimensional scale, in contrast to "semantic spaces" derived through factor analytic techniques, which constitute constrained multi- dimensional arrays of unidimensional scales. This alternative model rests upon the following assumptions: 1) Within a given cognitive domain, it is assumed that there exists a structure; i.e., a formal set of relations among the linguis- tic units used to describe objects residing in the domain. 2) It is assumed that the meaning of a linguistic unit is deter- mined by its dissimilarity relations (physical separation in the Spa- tial representation) to all other concepts in the domain. 3) Within a given domain, it is assumed that a subset of linguis- tic units will bear gtgblg_relations to each other, determined by cultural usage, describing a structure which is generally applied to other lin- guistic units representing objects within the domain. The subset of linguistic units so designated (e.g., adjectives) may be identified as having meanings (locations) determinable by reference to other linguis- tic elements of the subset, and independent of particular perceivable ref whit terr refe of di referents (objects of the domain) which might exemplify instances to which they refer. Two implications of this assumption of abstract de- terminability (without necessary reference to particular perceivable referents) are: a) that the relationships between elements of this subset will be as stable across time as the language of which they are a part; and, b) that the stable structural array of the subset will con- stitute a common, stable sub—structure in the individual cognitive structures of users of the language. 4) It is assumed that §§ can be taught to report ratio judgments of dissimilarities among traits and concepts. Within the semantic space characterized by these assumptions, it is useful to Specify definitions for a number of terms. An attribute will refer to a line segment between points representing linguistic units which §§ perceive as semantic opposites. Dimension refers to a refer- ence line, orthogonal to all other dimensions, through the configuration of attribute end points. Note that the goal of factor analysis has been to identify attributes which load highly on one dimension, but not on others. To designate this condition, one can say that for a given dimen- sion there may be an attribute or set of attributes that are exemplars of that dimension. Of course, there may also be any number of attributes which are not exemplars of any dimension. Tyipcally, non-exemplary attributes are purged from the interpretation of factor analytic solutions because they are not considered to be identifiably useful in the interpre- tation of dimensions of judgment in the domain. si by is 65 many be w one confc ists I ”1' 99 Posed bveen figur 03900, belng evdIUe mEanlr and MC model 10 However, a different logic operates in the analysis of multidimen- sional scaling configurations. Such configurations may be interpreted by use of property vectors or by projections of stimuli on axes. What is important is that the set of points be arrayed in as many dimensions as are empirically reliable. If a stable attribute is non-exemplar (located only moderately "high”) in the reliable dimensions, discarding the trait means that one is discarding highly reliable information. In the model proposed here, attributions of non-exemplar traits are con- sidered to provide useful information about probable attributions of many other (exemplar) traits, and are therefore retained. Before discussing the general assumptions of the model, it would be worthwhile to clarify assumption 3. Recall that in factor analysis one can sum across Ss, across concepts or across both; thus eliminating confounding variance due to §§ or due to concepts. Evidence clearly ex- ists which documents individual differences in the semantic space (Wiggins and Fishbein, 1969; Talbot, 1969). However, the model pro- posed here is concerned primarily with assessing the relationships be- tween linguistic units at a cultural level and the perception of public figures from the perspective of the aggregate. The effects of variance due to concepts is potentially problematic. Osgood §t_al, (1957) concluded that the nature of the concept being rated will influence the factor structure obtained: "...the more evaluative or emotionally loaded the concept being judged, the more the meaning of all scales Shift toward evaluative connotation" (p. 187). Ad- ditional research by Green and Goldfried (1965) and Rosenbaum, Rosenbaum and McGinnies (1971) further documents concept effects. The proposed model does eliminate concept effects that may be artifacts of ratings on 11 unidimensional scales which are factor analyzed because the structure of the semantic space is defined by the relationship between all bipolar adjectives, as opposed to allowing the means and variances of concept ratings to define the structure of the space. Thus, in the proposed model, a concept of highly emotional connotation will not influence the meaning of all scales, and, ultimately, the factor structure. Of course, concept effects cannot be totally eliminated because some of the impact of changing from concept domain to concept domain elicits true, non-artifactual changes in the structure. It is for this reason, as will become clear below, that the domain specificity assump- tion has been made. It should be noted that while the majority of research on "implicit personality theory" supports assumptions 1 and 3 of the model (see next section), two studies (Hanna and Jones, 1973; Doherty, 1973) found changes in the structural array of traits by changing the individual or concept being evaluated. The obtained changes did ggt include order changes of concepts in the structure. Indeed, cannonical correlations were quite high, .989 and .881 for a two factor structure obtained by Hanna and Jones (1973). The exact nature of the changes in the semantic structure when §§ were required to evaluate different "reference persons" dealt with changes in the distances between some of the adjectives. When changing concepts such as "hypocrite," "astronaut," "surgeon" and "killer," the obtained changes in the structure were expansions or con- tractions of distances between attribute end points that are relevant or irrelevant to the concept being evaluated. Doherty's (1973) results and discussion implied that "adequate" and “capable" were further apart in the "hypocrite" and "killer'I structures than in the "astronaut" and 12 "surgeon" structures. Further, “cruel" and "kind" were closer together in the "astronaut“ and "surgeon“ structures than in the other two. Doherty (1973) concluded that "...when the multidimensional scaling solutions are compared fer different references, they appear to be very similar. However, systematic changes may be induced, resulting in a change in the relative length of the capability dimensions fbr one of the negative references“ (p. 78). Thus, while the order of adjectives in the structure are similar, some variations in the distances between some concepts are obtained. (Unfortunately, Doherty did not have any independent criteria for demonstrating that attribute relevance or irre- levance is the explanatory variable for the expansion or contraction of attributes.) Note that the "reference persons" used in these studies are not people but are terms that constitute classes of peOple. Assuming that the domain "class of all individuals" is too broad a domain to be used to avoid concept effects, one may wish to break "domain“ into a hierar- chical set of domains of others. Therefbre, it would be advisable to provide the following definition: A cognitive ggmaig_is a set of objects or concepts that naturalistically possess some classificatory character- istic in common. (This definition, admittedly poor, seeks to avoid any uninterpretable or potentially confusing definitions, such as defining domain as "phenomenal objects which the person treats as functionally equivalent" (Scott, 1969, p. 262).) In a hierarchical clustering analo- gy, a domain at one level may be the set of all human beings. At anoth- er level, the set of personal acquaintances, or the set of current Amer- ican politicians. Within each domain, attributes will vary in terms of relevance and irrelevance. Thus, as specified in assumption 3, 13 stable structures should occur within domains, and there will be varia- tions in lengths of attributes from domain to domain. On the Assumptions of the Proposed Model Obviously, one would like to raise the question as to whether the above four assumptions, upon which the model rests, are tenable. Evi- dence concerning the ability to make ratio judgments of separation Specifically (assumption 4) is scarce. The most recent evidence con- cerning individuals‘ abilities to use metric MDS was presented by Gordon (1976). Gordon found that varying the criterion pair across nine inde- pendent samples produced statistically identical structures. Gordon, however, cautioned against generalizing these findings to data sets where SS were required to make distance estimates among heterogeneous concepts. In the present study, effort was made (in accordance with the assumption concerning domain specificity) to use a homogeneous set of concepts. More will be said about this assumption in the section on "Derivation of the Multiple Attribute Measurement Model." Fortunately, a plethora of research can be referenced in relation to the first three assumptions. For example, assumption 2 is commonly made in the MOS literature (see Shepard gt al., 1972), as well as in the research on the analysis of meaning (Miller, 1969). As such, this assumption needs no further articulation and support here. However, it would be profitable to review additional literature relevant to assump- tions 1 and 3. This research literature has typically been subsumed un- der the category of "implicit personality theory“ research. The next three subsections will define this construct, present research evidence which bears upon the generality by which "implicit personality theory" It) I/‘ dis perc PEOP whic has all situa all s of a able past estat Struc trait Iicati togetr e"l‘or‘ IASch. ality amPles :4 rESPOn I4 is applied by individuals as they perceive and evaluate others, and discuss questions which pertain to the nature of the phenomenon. (a) Definition of “Implicit Personality Theory" Some of the early conclusions of person perception research (Hastorf gt 31., 1958) were (1) people use a rather limited number of perceptual categories even when describing very different kinds of people; (2) there is a strong positive relationship between categories which people use in describing others and themselves; and (3) a person has both a core of generally consistent categories used in describing all people and a set of generally consistent categories which depend on situational factors. Further, a common, explicit assumption underlying all studies in person perception has been that the perceiver's judgments of a stimulus person are a function of both (a) the information avail- able about the stimulus person's characteristics, and (b) the perceiver's past experience with people ~~experience which presumably leads to the establishment of the "implicit personality theory," which, in turn, structures the individual's judgments of others. "Implicit personality theory" iS the set of formal relations among trait adjectives. The history of research in person perception has rep- licated the common finding that a person expects certain traits to "go together." Research on the "halo effect“ (Thorndike, 1920), "logical error" (Newcomb, 1931), "trait implication" (Hays, 1958) and "centrality" (Asch, 1946; Kelly, 1950; Wishner, 1960), as well as the "implicit person- ality theory" (Cronbach, 1955, 1958; Rosenberg and Sedlak, 1972) are ex- amples of investigations into the process of trait co-occurrence. The “implicit personality theory“ concept was first introduced as a response bias in accuracy scores (Bruner and Tagiuri, 1954), and a more A /_- 15 general statement was provided by Cronbach (1955) and Secord and Berscheid (1962). Cronbach noted that the rater's bias deserved atten- tion in its own right, beyond that of a source of constant error, and suggested that a judge's implicit personality theory could be described by the means, variances and covariances of the judge's ratings of a large number of others. Only a few studies have used Cronbach's opera- tional definition (Crow and Hammond, 1957; Gross, 1961). Gross found some evidence for bias in means and variability in ratings of 30 heter- ogeneous others — each presented under conditions of minimal information transmission (30-second films of each person at a park bench). However, the response bias accounted fer a negligible portion of the variance while stimulus factors, in spite of the limited infbrmation available, accounted for the major portion of the variance. Koltuv (1962) criticized the Gross study because the rating scales were few, and did not represent relevant dimensions in perceiving others - “This method of choosing dimensions for the perception of others may partially explain the finding that perceiver predispositions account for little of the variance in social perception..." (p. 5). Nonetheless, while intrajudge consistency of means across scales was fbund, Gross felt that the "generalized other" had little validity and that "...cul- tural similarity in the experience of the judges resulted in their drawing upon commonly held stereotypes" (p. 608). Bruner, Shapiro and Tagiuri (1958), as well as Hays (1958), have used the term "implicit personality theory" in reference to the network of relations among personality traits. According to Bruner gt gl,, a certain set of traits can be input into a matrix of "lay personality theory" from which other traits can be predicted. Their research on l6 trait combination was based on the following presupposition: "The fact of consistency of behavior, the backbone of personality theories, is represented in language by which people are commonly described. It is characteristic of trait words like honest, brave or clever that they do more than denote Specific acts of a person; that, indeed, they summarize or "package" certain consistencies of behavior" (p. 278). Hays (1958) presented a Similar account: "...a person mg§t_have some relatively stable scheme of expectations and anticipations about others.... This scheme may be thought of as a set of inferential rela- tionships among experienced attitudes and traits which exist for the in- dividual" (p. 289). He recommended two models for describing the fbrmal relations among traits, the implication model and the similarities model. The implication model assumes that when an individual infers one trait from another that such inferences are never made with absolute certainty. Data can be collected by informing subjects that the target individual possesses some trait x, and then the subjects are asked to judge the probability that the target possesses traits y, 2, etc., on a scale from 0 to probability of 1.0. Hays also postulated weights. However, what is important about the model is that the data is transformed from the origi- nal likelihood estimates to "true probability" estimates. Also notewor- thy is that the pairs of traits are only rank ordered and then subjected to Coombs' (1950) multidimensional unfolding procedure. Hays only used eight traits and did not label the dimensions obtained because of the small number of traits used. The second model, the similarities model, has not generated much research and will not be discussed. Todd and Rappoport (1964) compared the implication model with two sets of factor analytic ratings fOr Ss' ratings for real persons l7 and hypothetical others (i.e., "intelligent person"). The real persons were familiar person-objects utilized in a modification of Kelly's (1955) Rep Test. The comparisons indicated that the three procedures were in high agreement with respect to the degree of implication indi- cated between traits. The differences between methods were as expected. The two factor analytic procedures were in greater correSpondence with each other, and Hays' corresponded better with the ratings of hypothet- ical others than to real persons. However, Todd and Rappoport identified three problems with the im- plication model: there exists no analytic criteria to limit the number of dimensions to be extracted; there exists no criteria for determining the relative importance of the dimensions obtained, and, no convention exists for deciding what constitutes "significant“ loadings on dimensions. The more crucial finding was the factor analytic and implication model provided differences in terms of the number of dimensions obtained and Todd and Rappoport concluded that neither of the models provides satis- factory dimensions of cognitive structure. They recommended a more sophisticated procedure (Shepard, 1962), despite the fact that even today some of the problems with the implication model are unresolved issues in the area of multidimensional scaling. Thus, "implicit personality theory" is defined as a stable structure of the interrelatedness of attitudes and traits that are assumed to exist in others. Over time, after multiple experiences with heterogeneous and multiple others in multiple and heterogeneous situations, people build up certain expectations of what traits "go together" in others. These expectations are incorporated into the language peOple use to describe others. Hence, there is some overlap between shared, common experiences 18 that determine one's "implicit personality theory" and one's own indi- vidualistic experiences. For this reason, Gross (1960) attributed the obtained response bias to the fact that §§ drew upon commonly held stereotypes when rating heterogeneous others. A number of models have been developed to measure the fbrmal perceived relations among traits, culminating in Todd and Rappoport's (1964) recommendation that multidim- ensional scaling be utilized. It should be pointed out that while Hays and Bruner gt_gl, presented the above definitions of "implicit person- ality theory" their analysis fell short of adequately representing any complex structure, or of assessing the stability of such a structure. Bruner gt gl, demonstrated that the kinds of inferences §§ made from Single trait-names yielded an accurate prediction of the kinds of infer- ences drawn from combinations of trait-names, but no assessment was made of structure ngng, Hays' (1958) investigation of structure was limited to only eight traits, but was suggestive. Wishner (1960) and Koltuv (1962) presented the first studies that explicitly investigated structure. Wishner (1960) questioned the metho- dology by which "central" traits were investigated (Asch, 1956; Kelly, 1950) by illustrating that any trait on the stimulus list may be central by appropriate manipulation of the items on the check list or rating scale. The issue of "central" traits is not resolved, but Wishner dem- onstrated the difficulty in methodology. Nonetheless: ...the most important feature of Wishner's analysis is that he has provided us with a working model of the “implicit personality theory.“ It is simply a correlation matrix among traits, a matrix we all carry around with us. Each of us has an idea of what traits are closely related to each other. (Hastorf, Schneider, and Polekfa, 1970, p. 41) 19 Similarly, Koltuv (1962) conceptualized "implicit personality theory" as a pattern of nonzero intercorrelations which people assume to exist between traits in others. She demonstrated that this pattern remains nonzero when the halo effect is controlled through partial correlation. In sum, "implicit personality theory" is defined as a stable net- work of relations among traits (and probably other categories and attributes) that (1) function for the individual to summarize or char- acterize the behaviors of others, and (2) to enable the individual to anticipate the future behaviors of others. Further, evidence exists that indicates that "implicit personality theory“ structures our recall of others. D'Andrade (1970, cited in Schneider, 1973) had §§ rate the meaning of traits, rate a person immediately after interaction, and rate the person sometime after interacting. The results indicated that the matrix of recalled trait intercorrelations more strongly corresponded to the matrix of meaning trait intercorrelations than with the matrix of trait intercorrelations of the ratings of the person immediately after interaction. Thus, "implicit personality theory" is a determinant of our recall and description of others. (b) Generality of "Implicit Personality Theory" Demonstrating that such a structure exists and that it is a deter- minant of a person's descriptions of others is a good first step; how- ever, the structure will be of value only if generality can be demon- strated - that is, only if the formal relations among traits are applied for different categories of persons such as men-women, blacks-whites, teachers-students, etc. Secord and Berscheid (1962) addressed this question. They asked whether strong affect toward the stimulus person 20 being judged would change the perceptual processes - that the biases of "implicit personality theory“ take a different form for those person- concepts of high or low affect. Secord and Berscheid had each §_generate traits which s/he felt as belonging to Blacks and to Whites. §§ were then required to rate the probability of co-occurrence of stereotype and non-stereotype traits for White and Black stimulus persons. While some Shift in mean ratings occurred because of categorization of a person as Black, the associations between stimulus traits and judged traits remained remarkably consistent whether the stimulus person was Black or White. They argued that "...the concept of implicit personality theory may be presumed to have survived this relatively stringent test of generality" (p. 77). Additional support for the generality of "implicit personality theory“ was offered by Koltuv (1962), who feund that trait intercorrela- tions among traits were stronger for unfamiliar acquaintances than for more familiar ones. Thus, differences in "implicit personality theory" fbr close or distant acquaintances differ only in degree, not kind. Passini and Norman (1966) found high factor loadings for close friends and lower factor loadings for strangers, but the factor structure re- mained the same. This latter study is highly suggestive, since it indi- cates that people not only carry around a matrix of trait intercorrela- tions that applies to acquaintances, but that it also applies to strang- ers. Whether such ratings had any validity m-any effect on subsequent behavior --is another question. (Indeed, there was little correspondence between self-reports and ratings by others.) These results indicate that pe0ple tend to assume that a trait x is jg general associated with a trait y. (See also Jones and Nisbett, 1971.) 21 In general, the results of the above studies clearly support the robustness of the formal relations among traits. Similar to Koltuv (1962), it appears safe to conclude that changes in reference persons result only in subtle changes in degree of perceived co-occurrence, and not in the change in the structure itself. (c) The Nature of the Phenomena The generality issue is highly related to the issue of the nature of "implicit personality theory." The traditional explanation for the existence of "implicit personality theory," expressed by Bruner gt.gl. (1958), is that the individual has many different types of experiences with many different types of persons and based on these experiences the individual learns what traits "go together." The results of the Passini and Norman (1966) study, that a similar factor structure was obtained for close acquaintances and for strangers jg_virtual absence 9: prior acquaintance demonstrated that the dimensions of perceiving others rest implicitly in the perceiver and are (presumably) activated with very superficial information and observable cues. They argued that the "im- plicit personality theory" operated as the basis by which raters arrived at nearly consensual judgments of strangers and that increased acquain- tance with the raters increased the loadings on these factors. Muliak (1964) and D'Andrade (1965) offered strong criticisms to the position that raters learned from experience how traits go together in others. Instead, they argued that the "implicit personality theory" rep- resents the relationship between trait adjectives according to the mean- ing of the words and not according to how the traits co-occur in others. Muliak developed a trait-rating instrument using 76 trait adjectives I L1,... _2 22 (from an original pool of 200). Three sets of SS rated, in three separ- ate studies: (1) 20 personalities -- 10 famous persons and 10 persons the SS knew; (2) 20 stereotypes (ex., "intelligent person"); (3) the mean- ing of 20 traits -- traits which were randomly selected from the list of 200. Summing across raters and things rated, each matrix of intercorre- lations was factor analyzed. The ratings of real people resulted in eleven factors; of stereo- types, ten factors; for the meaning of trait words, nine factors. Con- sidering that the 20 words were chosen gt_random, it is surprising that three factors from the study of the meaning of trait words showed con- gruent relationships ("similarity coefficients" greater than .79) to four factors in each of the two other studies. These three factors, and the two sets of four from each of the other studies, accounted for 60% of the common variance in their respective studies. These results suggest that it is not necessary to rate actual people in order to determine the "personality factors" that would be associated with a set of trait words. The typical conceptualization of implicit personality theory holds that the raters have learned from ex- perience which traits go together in actual persons, and that this “packaging" or summarization of the generalized other is represented in the factor structure. Muliak (1964) argued against this: This is a pertinent objection in the case of ratings of stereotypes. But it seems to require accepting many assumptions without evidence in the case of the study of ratings of the meanings of trait words. The SS of this study were not asked to rate the traits on the degree to which traits went together in persons. They were asked simply to rate the trait words at the top of the rating scale according to how close they were to one or the other poles of the bipolar trait-ratings scales in meaning. It was 23 assumed therefore that the 55 did what they were asked to do. But the author would be willing to consider the above objection as valid if someone would produce evidence that raters of the meaning of trait words make such ratings according to their knowledge of how traits go together in persons and not according to their knowledge of meanings as such. (pp. 509-510) D'Andrade (1965) further supported the linguistic explanation: ...the hypothesis proposed here is that correlations and factors obtained in Norman's study are derived be- cause sets of these terms partially overlap in meaning. This type of partial overlap in meaning appears to be a general linguistic phenomena, resulting from the fact that Inost lexical items in a language are com- posed of a cluster or bundle of meanings which recom- bine in sets to form different words. The meaning units which compose such bundles may be referred to in linguistics as "sememes" or sememic components.... (pp. 216-217) ...From this point of view, the meaning of words are composed of a bundle of dimen- sional values. (p. 222) D'Andrade had 10 55 rate all paired comparisons of Norman's Pole 8 traits on a seven point scale of Similarity of meaning. Though the data is appropriate for multidimensional scaling, D'Andrade performed a fac- tor analysis. He obtained a five factor structure that was highly Sim- ilar to that of Norman (1963) and Passini and Norman (1966). Only five factor loadings were misplaced. This controversy, however, has lost its impact since it is not clear how the underlying processes of judgments of similarity of meaning and judgments of perceived trait covariations are separable. The Simil- arity of meaning hypothesis is inconceivable without a foundation in per- ceived trait covariation or implication since language itself is asso- ciatively and experientially determined. For example, Friendly and Glucksberg (1970) offer some insight into how new linguistic items are incorporated into the semantic Space. At Princeton there exists a Spec- ific student slang. Friendly and Glucksberg had freshmen and seniors 24 sort both slang terms and adjectives. Their results, which indicated a two-dimensional configuration for freshmen and three dimensions fOr seniors, is compatible with "the notion that the acquisition of a spec- ific subcultural lexicon involves, at least in part, the acquisition of semantic dimensions relevant to the Specific values of that sub-culture" (p. 59). Further, the seniors differentiated more along the slang terms. Friendly and Glucksberg (1970) asserted that: "In order to use the terms appropriately, it is necessary to learn which attributes of their referents are critical, for example, what distinguishes between 'wonk' and 'non-wonk'" (p. 63). Hence, it can be argued that as one learns to use labels (traits) of a language one necessarily learns to differentiate along the attri- butes relevant for the sub-culture or culture. Generalizing such find- ings to a cultural level, one would expect strong consensus among the relations among traits given the condition that there exists a consensual nature of meaning of the language users (Wittgenstein, 1953; Barnett, 1975). In sum, "implicit personality theory" is a general cultural phenom- emon expressed in the normative use of language. The dispute over the nature of "implicit personality theory" is artificial because if meanings of trait labels (and the formal relations among traits) were not congru- ent with the way traits are perceived by the individual as covarying in actual others, then the meanings of traits as such would change. Note that Bruner gt_gl, (1958), in presenting the "realist“ position, asserted that consistency of behavior is incorporated into the language by which people are commonly described. While "consistency of behavior" is a problematic assumption, especially after Mischel's (1968) work on the 25 relations between personality tests and behavior, one can at least argue that people perceive more consistencies in the relations among traits because of informational biases; people see x types of people only in y types of situations. Hence, perceived consistencies are maintained. In sum, it would appear that the assumptions of the model are fair- ly well supported assumptions. Two qualifications, however, are in order. First, a good deal more research must be conducted on the ques- tion of individuals' abilities to make ratio judgments. Second, it is obviously the case that the stability of the array of traits in the semantic Space will be affected by the number of person-concepts which are included in the analysis. We have argued that the location of each concept in the multidimensional space depends upon its similarities and dissimilarities with all other concepts in the space. If most concepts are traits, then there would be an excellent chance that one is tapping the true relationship between the set of traits sampled from the ”impli- cit personality theory" because the location of each trait is "anchored" primarily by its assessment with other traits. However, in the case where only a few traits are included, the location of each trait would be determined by person-concepts. It is not known at this time how 10- cating traits into a space based on distances from person-concepts will affect the stability of the relationship between traits and is one avenue for future research. {2' ,-,_J‘ 26 Derivation of Multiple Attribute Model In this section*, we shall give a general description of the deri- vation and characteristics of the proposed model. The approach to ratio scaling in communication research (see Gilham and Woelfel, 1975, for review) is based on Einstein's conception of the measurement of distance: For this purpose (the measurement of distance) we require a "distance" (Rod 5) which is to be used once and for all, and which we employ as a standard measure. If, now, A and B are two points on a rigid body, we can construct the line joining them according to the rules of geometry; then, starting from A, we can mark off the distance S time after time until we reach 8. The number of these opera- tions required is the numerical measure of the dis- tance AB. This is the basis of all measurement of length. (Einstein, 1961, p. 6) The analogous measurement procedure proposed here is two-staged: first, an arbitrary distance (or dissimilarity in the general case) is stipulated between two elements of the stable subset of linguistic ele- ments constituting a part of the language Spoken by S. It is vital to note that rules for the perception or measurement of this initial mea- surement distance or discrepancy are not stated; rather, the scientist must assume the subject and himself/herself share a common referent for the ordinary language symbol "distance" or “difference," and that the subject can make this initial recognition unaided by further definition. Ultimately, it is an g_priori call to common experience as codified in ordinary language symbols that establishes a link between the everyday experience of the observer and the scientific theory. Secondly, the scientist Specifies a rule by which other instances of distance or dissimilarity are to be compared to this unit. In this * Much of this discussion also appears in Cody gt 91,, 1976. 27 case, the observer is asked to make ratio comparisons of all other dis- tances or discrepancies to this arbitrary standard. Since this technique yields both a true zero (that is, no differ- ence between two stimuli) and a standard unit or interval of measure (Rod 5), it may be seen to constitute, by definition, a ratio scale whose validity rests on the conventional linguistic symbol system. This means that numbers yielded by these procedures represent discrepancies among stimuli as they appear to the respondents, rather than as defined by the scientist. Formally, when these procedures are performed for a Single observer over the (N(N-l))/2 possible non-redundant pairs of N stimuli, they yield a symmetric matrix S'where any cell Sij represents the discrepancy or difference between the ith and jth stimuli as re- ported by the observer, expressed as a ratio to an arbitrary discrepancy Sxy (Rod S). Techniques which map the structure of discrepancy or dissimilarity data onto a Space where it may be interpreted as distances are well known in the multidimensional scaling literature, and have been since Torgerson (1958) defined the procedure. Computational equations for Torgerson's method, called metric, classical or Torgerson multidimen- sional scaling, have been detailed in several places (Torgerson, 1958; Woelfel, 1974; Serota, 1974) but certain salient aspects deserve mention here. First, metric multidimensional scaling (MMDS) yields a coordinate system of 5's (N-l) orthogonal dimensions for N stimuli. Second, the mapping of discrepancies into this space is one-to-one; that is, no in- formation is lost by MMDS. Third, the function which maps discrepancies (Sij) reported by the reSpondents onto distances in the Space (S'ij) is bi Ka «it al lii Iar 28 the Simple that is, distances in the Space conform exactly to discrepancies reported by the respondent(s). The latter two of these characteristics do not hold for non-metric models (Kruskal, l964a,b; Young and Torgerson, 1967; Lingoes, 1972). Proponents of non—metric multidimensional scaling gener- ally reject the metric model on the basis of the following two assump- tions. First, many psychometricians, for philosophical or heuristic rea- sons, resist the notion that k, the dimensionality of the space, should be left a free parameter to be discovered inductively as a consequence of the rule for measuring distances (Shepard, 1972; Veldman, 1974; Kaiser, 1958). Rather, they feel that k_shou1d be set at some arti- trary small value and distances (dissimilarities) reported by observers adjusted accordingly. The view taken here, however, is that the gener- ality of language and the applicability of many linguistic units to a wide variety of concepts and contexts make it at least plausible that linguistically-determined semantic Spaces be represented as having a large number of dimensions. This plausibility alone is sufficient reason to reject arbitrary constraint of dimensionality, leaving the question of dimensionality to empirical resolution. Second, respondents are generally assumed under non-metric models to be unable to make reliable ratio judgments of discrepancies among stimuli (Coombs, 1964; Shepard, l962a,b). It is assumed here that re- Spondents ggg_make such judgments. This assumption is supported by re- cent empirical evidence that most of the apparent unreliability in [O 29 individual judgments may be systematically explained in terms of indi- vidual self-perception and cognitive processes; i.e., by individual per- spectives, or points of view, within a culturally-normative domain with- in which the arrangement of stable concepts is determined by an aggre- gate of which the individual is a member (Marlier, 1974). In addition to the general characteristics of MMDS spaces noted above, three others are particularly relevant to the comparison of the proposed scale and factor analytic models. First, no assumptions are made in the MMDS Space about the semantic meaningfulness of the centroid. Consequently, no assumptions need be, or are, made as to attribute end point equidistance from, or bipolarity with regard to, the origin. Attributes are not constrained to intersect at a common point (which is selected mathematically but nay not accurately represent subject percep- tions of the relationship of attributes as they occur unrestrained), and stimuli which are not perceived by respondents to project on an attri- bute are not constrained to do so. Therefore, ceiling effects are elim- inated. Second, the mapping of dissimilarities represents an example of fundamental ratio measurement, and no standardization is involved in the MMDS routine. As a result, attribute lengths and differentiation are not imposed by the researcher for mathematical rather than theoretic reasons, but may be represented as expressed by respondents. The result is high precision of scaling and increases in absolute amounts of reliable var- iance in scaled perceptions of stimuli (Danes and Woelfel, 1975). Third, attributes in the space need not be exemplars of any dimension. Inter- pretability of the MMDS space rests, in fact, on the distances of scaled stimuli from the trait adjectives which constitute the scale. Conse- quently, purging of non-exemplary attributes, which has the effect of i O z" .' 30 reducing the total spatial volume near semantically meaningful points, reduces interpretability, and is not called for. Unlike factor analytic representations, which seek simplicity of representation through divi- sion into mathematically independent parts, the MMDS semantic space seeks an accurate and theoretically useful representation of interdepen- ggggg, Comparability of MMDS Spaces across administrations depends not on the orthogonality of semantically meaningful axes, as with fac- tor analytic Spaces, but on the stability of the configuration of de- scriptors in the aggregate Space. Rotation of aggregate Spaces to a least-squares best-fit of theoretically expected stable concepts (Woelfel gt gl,, 1975) has been shown empirically to yield highly stable configurations (Danes and Woelfel, 1975), thus establishing the compar- ability of scales of the type proposed here. Application of the MMDS scale to measure individuals' perceptions of stimulus attributes involves the generation of semantic spaces for in- dividual respondents in which the aggregate configuration of stable de- scriptors is maintained. Thus, a scale generated from the aggregate NxN matrix (Sij) may be applied to M stimuli by requiring respondents to apply the arbitrary standard dissimilarity (Sxy) in making ratio judg- ments of the dissimilarity between all possible pairs of the M stimuli, and between each of the M stimuli and each of the N descriptors. This procedure generates a new (N+M)x(N+M) dissimilarity matrix Sij*' The space generated from this supermatrix represents the respondent's percep- tion of stimuli (objects of the domain) relative to semantically meaning- ful points which the respondent (or any speaker of the language) might use to describe the stimuli. The location of any stimulus in such a ~82 ilk fix 31 ,DETBRMINED INDUSTRIOUS o .SKILLI-‘UL 'INTELLIGENT (P) serious COLD l\ important ' \ :“mcmble ‘ CAUTIOU/ oPRACTICAL umorless I unpopular ’rrit--le ‘ unhappy / reliable ORI 4) 454a!:. honest ; Vh\ dishonest ‘-.i§~\\ insig ificant unreliable humorou\‘ hapEyUl good-nature- \\ P p.a:: frivolous WA SOCI e Figure l. Hypothetical illustration of Multiple Attribute "Scale”. From Rosenberg, Nelson, and Vivekananthan (1968). ""i _ > 1 tr to St co in of pm spa to ( ti 32 space therefore represents the "meaning" of that stimulus for the res- pondent, represented in a quantifiable relationship to known points whose meaning is Shared by the respondent and other speakers of the language. In the semantic Space generated by procedures suggested above, the result would be a "scale" as represented in Figure l, in which a stimu- lus person (P) has been located relative to the stable configuration of trait descriptors. Interpretation of respondent attributions of traits to stimuli as located in the Space would appear at first glance to be straightforward. If an attribute is represented as the line segment connecting the linguistic units which would bound a unidimensional Scale for the measurement of that attribute or pr0perty, then the scaled value of that attribute in the MMDS representation would be determined by the point at which the stimulus projected onto the attribute in the MMDS space. Thus, in Figure 2, the relative amount of "goodness" attributed to an object 0 would be given by the difference in distances between Op (the projection of S on the good-bad attribute) and good, and between nd b d. Op a a Q ...1. Good Bad p Figure 2 Such an interpretation follows traditional utilization of factor analytic Spaces, in which stimuli are located such that their projections 33 on an arrangement of orthogonal vectors correspond to unidimensionally- scaled values of those properties for the stimulus. Since all distances in the MMDS space are ratios of the standard dissimilarity (Sxy)’ quan- tifications of attributions in this manner are continuous, and therefore represent an increase in precision over the ordinal or assumed interval levels of measurement typically achieved in factor analytic Spaces. In the semantic Space generated by procedures suggested above, how- ever, a configuration such as that illustrated in Figure 3 (in two dim- ensions for illustrative clarity) is also possible. G od ~ - Bad 0 [_‘ Qp angerous 0“ Figure 3 In this hypothetical example, the reSpondent's attribution of "goodness“ to stimulus Q could be quantified through the procedure dis- cussed above, resulting in a neutral value. Similarly, we could expect from the configuration that the respondent would scale 0 at the "danger- ous" end of a semantic differential scale anchored by the adjectives "dangerous" and "safe." Since 0 does not project onto either the "active-passive" or "hard-soft" attributes, however, quantification of 34 the respondent's attribution of these properties to Q is not possible by the procedures discussed above. Two interpretations are possible, both of which may be plausibly illustrated if we assume for the moment that Q is a gun. In this instance, a respondent might well place an X at the "hard" end of a "hard-soft" semantic differential scale, since "hardness" is an obvious, if unimportant, property of a gun. Asked to scale 0 (the gun) on an "active-passive" scale, however, the respondent might well be stymied by the conflicting perceptions of actual passivity and potential activity. Faced with this ambiguous perception, such a respondent might well decide that the "active-passive" continuum is irre- levant to his primary perception that the gun is dangerous, and thus mark the neutral point in the semantic differential to indicate his per- ception that the scale is inapplicable. The point, of course, is that neither hardness nor activity are salient attributes in the respondent's perception of the gun. A semantic space generated through factor analysis would fail to represent this lack of salience. In such a space, as noted previously, all concepts are constrained to project on all attributes (exemplary or non-exemplary) which are constrained to intersect at a semantically meaningful origin. But the example above illustrates the ambiguity of the origin's "meaning," and the constraint that every concept must pro- ject on every standardized attribute makes differentiation of salient from non-salient attributes impossible. Consequently, a factor analytic representation of the example above would either represent the correla- tion between "active-passive" and "good-bad" as artificially high (if Scaled perceptions of the gun were submitted to factor analysis), or re- sult in an indeterminate location of Q (the gun) in the semantic space 35 (if the arrangement of attributes had been determined previously by factor analysis of scaled perceptions of other stimuli). In the MMDS model, however, the ambiguity of interpretation of the example is resolvable. A ratio measure of respondent attribution of activity-passivity to Q, for example, is given by the difference in the distance from the stimulus O to "active," and Q to "passive." The range of possible values of this measure is t the length of the attribute in the space. A value of zero indicates neutrality, and occurs when S has a projection onto the midpoint of the attribute. A ratio measure of the salience of an attribute to the respondent's perception of a stimulus may be obtained by subtracting the distance from the stimulus to the attribute from some arbitrarily large constant. In the case where a stimulus can be projected onto an attribute, the distance between them is the distance between the stimulus (Q) and its point of projection (Op). Where projection is impossible, as in the above example, the dis- tance between the stimulus and the nearest end point of the attribute is the distance between 0 and the attribute. Thus, in Figure 4, the quanti- fied attribution of "goodness" to stimulus O1 is given by (a-b). The salience of the good-bad attribute to the respondent's perception of Q1 is (k-c), where k is any large constant. Similarly, the "goodness" of 02 is given by (d-e), and the salience of the good-bad attribute to per- ceptions of'Q2 is (k-d). Q a,b,c,d,e, = lengths Q l of respective lines 2 a b as ratios of standard e c distances Sxy d Good e r—# Bad Q1P Figure 4 36 By salience is meant the degree to which a concept is defined by an attribute. Conceptually, it is the same as attribute prominance (Zajonc, 1969) or attribute relevance (Shrauger and Petterson, 1974). Prominance was defined by Zajonc as the ability of an attribute by itself to represent or characterize the referent (p. 329). He operationalized the concept in terms of rank orders of what attributes were most charac- teristic of the referent. Even the crude rank orders of attributes used as weights increased correlations between simple average of the indi- vidual attribute valences and overall attitude from (.22) to (.66) for the weighted average. Shrauger and Petterson (1974) obtained attribute salience measures for the "self" by having §§ select out of 57 attributes the ten which were "most relevant and important," and the ten which were “least rele- vant and important." Results clearly indicated that attributes which were highly relevant for the self were used more frequently in describ- ing others than non-relevant attributes. The importance of attribute salience cannot be understated. In addition to sets of attributes associated with a theoretical construct (i.e., credibility), the salience of each attribute is an indicant of the weight placed on that attribute. While a source may have a score on (either exemplar or non-exemplar) attributes w, x, y, and 2, one or more of those attributes may be totally irrelevant and non-salient in perceiving an object. In Figure l, the attributes of serious-frivolous, and important-insignificant are much more salient in the Ss' perceptions of P than reliable-unreliable and sociable-unsociable; yet P does have a score on all four of the attributes. O O. ul- 37 An assessment of attribute salience is critical when one considers the design of messages intended to manipulate credibility. First, note that there is perscriptive utility in using the concept "Ideal Credible Source" (McLaughlin, 1975; Heston, 1973). The location of this point in the space provides information concerning the desired level (or score) on each attribute such that the set of scores represent maximum credi- bility. Second, an assessment of salience for each attribute can be ob- tained. Once the salience of attributes in the perception of the "Ideal Credible Source" are determined, manipulations of these key attributes will result in the greatest amount of change towards (or away from) the location of the point representing "Ideal Credible Source." Previous research on the manipulation of source credibility, stem- ming from factor analytic research, has centered on manipulating one or more factors. Most illustrative of this point is the manipulation of expgrtise (Hovland, Janis and Kelly, 1953; Aronson and Golden, 1962). Such manipulations had significant impact on the degree of attitude change. Yet, there exists a problem in manipulating an absolute low credible source. The most carefully conducted research on this point is Greenberg and Miller (1966). In experiment I, the character of the source was attacked on the grounds of unethical business practices. This negatively valenced induction was somewhat less than successful - approx- imately one-third of the Ss in the Low-Credibility induction group rated the source as "quite trustworthy." After a series of experiments, the investigators concluded: Even though audience members were given information that should have prompted them to question severely the com- petence and trustworthiness of their sources, a number of respondents failed to rate the source's credibility low in any absolute sense. While this reluctance to re- spond negatively may have been partially due to the he! an 38 quality of the message, the investigators believe that some additional variable is involved. Specifically, as mentioned earlier, a normative standard may operate in such a manner that audience members give a source the benefit of a doubt (i.e., in the absence of personal experience with the source, audiences may respond to sources in a somewhat positive manner). However, several plausible alternative explanations are also oper- ative. First, instead of a "normative standard" there may exist a re- luctance to use the negative end points of the seven point scale. Sec- ond, the question can be raised as to the degree to which "character" as operationalized by Greenberg and Miller (1966) is salient to credibil- ity as operationalized as competence and trustworthiness. Further, it is not clear to what degree positive induction and negative induction messages have been comparable. For example, Kelman and Hovland (1953) attributed a persuasive speech to a respected judge (positive induction) and to a man who was described in such a way as to give "the impression of being an obnoxious, self-centered individual with a Shady past and present" (p. 329). Such credibility inductions clearly have had impact on the amount of attitude change obtained. How- ever, it is difficult to argue for a "normative standard" by which mem- bers of the audience give the source a benefit of a doubt when there are several alternative explanations. In the present analysis, the movement of a source (a "target" public figure) to a theoretical point is best accomplished by the design of a message that moves the target public figure through (potentially) several dimensions. A new technique proposed for political communication (Woel- fel, Fink, Holmes, Cody and Taylor, 1976) is directly applicable here. The technique provides the best solution for obtaining the shortest path between the location of the target in the Space and the desired location k _ ‘._.=J 39 ("Ideal Credible Source"). The procedure takes into consideration all bipolar end points in the space, and, based on vector addition, computes either single vectors or n-vector resultants for moving the target to the ideal point. Thus, the technique provides information concerning what attributes should be associated with the target, which should be disassociated and the degree to which each attribute should be weighted in the credibility induction message. Summary and Hypothesis In sum, the assumptions of the semantic differential, and factor analysis of semantic differentials, are weakly supported. A new measure- ment model has been proposed that is both more commensurate with scaling assumptions and does not restrict every concept to have a projection on every attribute. The new model also possesses pragmatic advantages in the measurement of saliency of attributes. This new measurement model is applied to the measurement of the credibility construct. As mentioned earlier, the model can assess wheth- er attributes used in the experiment are relevant (salient) attributes. Further, the location of all concepts (bipolar and points, public fig- ures and the concept “Ideal Credible Source") provides the necessary in- formation for maximum low and high credibility inductions. The following hypotheses will be tested: H1: The messages based on procedures discussed above can significantly alter the perceptions of a source's credibility. 2: Persuasive messages attributes to a public figure perceived as closer to the "Ideal Credible Source“ will stimulate more attitude change than messages attributed to a public figure who is perceived as more distant from the ideal source. In addition, the following question is raised: Given the conditions of 40 using the same public figure in both credibility inductions, and that the message used in the negative valenced credibility induction is the antithesis of that used in the positive valenced credibility induction, will movement toward the location of "Ideal Credible Source" be greater than the movement away from this point? That is, does there exist bias in the Ss' processing of information such that they give the pub- lic figure in the negative valenced induction the benefit of a doubt? CHAPTER II METHODS This study is drawn entirely from data collected in the Fall of 1975 and Winter of 1976. The methods used in data gathering and analysis will be discussed in the following order: 1. SS utilized in the present study, 2. A Sorting Task and Criterion Pair Selection, 3. Selection of Public Figures, 4. Topic Selection Pretest, 5. Message Design Pretest, and, 6. Posttest. Subjects Subjects were students enrolled in undergraduate Communication classes at a large Midwestern University. All §§ participated in the study on a voluntary basis and received course credit for their participation. A total of 343 §§ were employed in the following phases of the study: (a) a sorting task including ninety-six trait adjectives (n=l8); (b) Selection of Public Figures (n=54); (c) Topic Selection Pretest of atti- tude topic used in the manipulation check (n=33); (d) a Pretest for ob- taining reliable location of the public figures in the attribute space (n=54); and (e) the Posttest (n=l84). Since data analysis is to be conducted with data collected from the posttest groups and the pretest for message design group, it would be worthwhile to assess comparability across these groups. Table 1 presents demographic data on the §s utilized in these five groups. The mean 41 42 TABLE 1 Demographic Infbrmation on_§s Utilized in Pretest on Message Design and the Four Posttest Groups. Speech Pretest Control Only Positive Negative n 54 47 45 45 47 7 Age 20.63 19.48 19.11 19.20 19.46 Male 25 16 22 23 22 Female 24 27 24 22 24 Race: Caucasian 43 35 41 38 44 Black 4 6 3 4 2 Other 2 - l 3 1 Not Reported 5 6 - - - r h »E 43 age of the students used in the pretest (20.63) was only slightly higher than the mean ages of §§ in the four posttest groups (19.48, 19.11, 19.20 and 19.46 for the control, speech only, positive and negative groups, re- spectively). Males and females were equally divided into each of the groups except in the control group where females outnumbered males. The majority of §§ were Caucasian (86.4% in the posttest groups). Sorting Task and Criterion Pair Selectipn The ratio scaling procedure requires that one distance between two traits be given as a standard. The distance between the two selected traits Should meet the conditions that the traits (1) have fixed loca- tions relative to each other, which are not spatio-temporally bound; and (2) are perceived unambiguously by Ss-u-that is, there exists extremely low variability in their locations. Traits that met the above criteria were identified by a sorting method. Ninety-six traits were selected from McCroskey, Jensen and Todd (1972), Osgood, Suci and Tannenbaum (1957) and Walters and JackSon (1966). In late October, 1975, a sample of 18 §§ sorted these traits into eleven categories. The underlying dimension was specified as social desirabili- ty. To minimize ambiguity, the role "a close friend" was used. There- fore, one end of the eleven point scale was labeled, "A quality a close friend of mine gppp1g_possess." The other end was labeled "A quality a close friend should not possess." Each adjective was typed on a card. The instructions fbr sorting the 96 cards are included in the Appendix (see APPEOdIX). Subjects were given a copy of the instructions to read while the experimenter read the instructions to them. The purpose of the sorting task was to identify traits that are per- ceived unambiguously by §§ along the social desirability dimension. Means 44 and variances were computed for the placements of each of the traits. Table 2 presents the means, standard deviations and distances from the mid-point for the ninety-Six trait adjectives. The mid—point is the mid-point of the eleven categories §§ had in which to sort the adjectives. The first six pairs were used in the present study. These pairs of ad- jectives were selected on the basis of having low standard deviations. The two traits selected as the criterion pair were intelligent and inex- perienced. Selection of Public Figures The goals of the pretest for public figures were two-fold. First, it was necessary that the public figures be recognized by Ss. Second, it was desirable that the target public figure, whose credibility was to be manipulated, be moderately familiar to the §§ but not perceived as currently active by the Ssim-that is, the §§ were not currently obtain- ing information about the target. This was necessary to ensure that the credibility of the target public figure did not change due to history in the time between the pretest and the experimental manipulation. A list of nineteen public figures was drafted and on January 8, 1976 54 §§ responded to questions pertaining to familiarity, occupation, no- table activities, interpersonal sources of information about the public figure, and mass media sources of information. The familiarity scale was a Likert-type item. Occupation and notable activities were open-ended questions. The two information source questions essentially asked, "In the past two weeks, how many times have you heard from family, friends, acquaintances about ?“ 45 TABLE 2 RESULTS OF SORTING TASK Means, standard deviations and distances from the mid-point for ninety- six trait adjectives (n=l8). __ standard distance from Adjective X deviation midgpoint just 2.11 0.91 3.89 unjust 10.33 1.28 4.33 intelligent 2.17 0.98 3.84 unintelligent 9.50 1.42 3.50 reliable 1.66 1.28 4.34 unreliable 0.38 1.46 4.38 competent 3.03 1.35 2.95 incompetent 8.88 1.53 2.88 experienced 4.16 1.34 1.84 inexperienced 7.22 1.44 1.22 repulsive 0.50 1.04 4.50 attractive 4.11 1.99 1.89 responsible 2.05 1.30 3.95 unresponsible 0.05 1.21 4.05 informed 3.00 1.41 3.00 uninformed 9.27 1.40 3.27 kind 2.27 1.36 3.73 cruel 0.72 0.67 4.72 awful 0.16 0.98 4.16 nice 3.50 1.65 2.50 friendly 2.50 1.12 3.50 unfriendly 9.50 1.75 3.50 pleasant 2.55 1.72 3.45 unpleasant 9.94 1.21 3.94 believable 2.61 1.29 3.39 unbelievable 9.61 1.88 3.61 cheerful 3.05 1.30 2.95 gloomy 9.67 1.71 3.67 energetic 3.11 1.57 2.89 tired 8.16 1.79 2.16 'hx Itl a TABLE 2 (continued) Adjective good-natured irritable logical illogical intellectual narrow cautious adventurous sympathetic unsympathetic relaxed tense active passive honest dishonest confident lacks confidence expert inexpert trained untrained sociable unsociable warm cold reputable disreputable calm anxious regressive progressive talkative Silent kO-b \ON com com Nah CDN O-‘ woo com 500) N01 ow ON \ON NO 0500 VOL) .61 .22 .83 .OO .33 .22 .05 .55 .OO .05 .22 .16 .16 .55 .83 .67 .88 .77 .88 .83 .55 .44 .16 .78 .55 .11 .66 .27 .44 .55 .50 .77 .83 .66 46 standard deviation .33 .96 .46 .94 .91 .48 .86 .95 1.91 —-m an.) an N—I N—I am N-a N—I om N—a N—I -- .—l .86 .48 .20 .38 .38 .36 .67 .32 .10 .74 .25 .38 .38 .65 .16 .76 .42 .70 .81 .20 .72 .43 .73 .88 .38 distance from mid-point 3.39 3.22 .17 .OO .67 .22 .95 .45 .OO .05 .78 .16 .84 .55 -'N MN 0000 090 ON 0000 .17 .67 .12 .77 1.12 .83 moo Ab u—J .45 .44 .84 .78 .45 .11 woo MN MO .34 .27 .56 .55 ON 00d .50 .23 .17 .66 —"N coco I L f‘. L/ -..J 47 TABLE 2 (continued) __ standard distance from Agjective X deviation mid-point extroverted 3.66 1.88 2.17 introverted 7.83 2.61 1.83 good 3.05 1.92 2.95 bad 4.22 2.04 1.78 qualified 4.55 2.01 1.45 unqualified 8.50 1.95 2.50 excitable 5.05 2.29 0.95 composed 3.66 1.94 2.34 poised 3.94 1.98 2.06 nervous 7.78 2.13 1.78 ethical 3.16 1.98 2.84 unethical 8.55 2.91 2.55 bold 4.22 2.04 1.78 timid 7.72 2.61 1.72 impressive 4.44 2.50 1.56 unimpressive 8.33 2.25 2.33 outgoing 3.67 2.22 2.33 withdrawn 8.05 2.98 2.05 meek 7.44 2.89 1.44 aggressive 3.77 2.39 2.23 altruistic 8.39 2.59 2.39 self-centered 3.77 2.78 2.23 selfish 8.66 3.05 2.66 unselfish 3.16 1.89 2.84 emphatic 4.28 3.20 1.72 hesitant 7.22 2.02 1.22 polite 3.27 1.64 2.73 blunt 5.16 3.17 0.84 verbal 3.83 2.04 2.17 quiet 5.72 2.11 0.28 cooperative 3.05 1.63 2.95 competitive 4.55 2.57 1.44 ‘- Q - amzuw 43 Mean ratings on familiarity and importance are presented in Table 3. Table 4 presents the mean number of times §s heard about the public figure from (a) the Media and (b) Interpersonal sources. Table 5 presents data on the §s' ability to identify the occupation held by the public figures. Data pertaining to the “notable activities" question is not presented be- cause of low response rates and high variability in reporting activities (thus making it difficult to conveniently present all the activities re- ported). From this data, two highly familiar pub1ic figures who were not high- ly active in the §s' informational environment were chosen as individuals to be used in the M05 paired comparison questionnaire: H. Humphrey and G. McGovern. The target individual was selected on the grounds of recog- nizability, moderate familiarity and low activity in the infbrmational en- vironment. The target individual selected was Birch Bayh. Topic Selection Pretest The manipulation check on the effectiveness of the credibility induc- tions requires a persuasive message be attributed to the public figure. Thus, the goal of the topic selection pretest was to select a topic on which §s were in high agreement (so as to minimize within group variance when aggregated measures of attitude change were computed). On January 13, 1976, thirty-three §s were asked to indicate the extent to which they agreed or disagreed with each of eighteen attitude statements. E1even point Likert-type items were used for this purpose. Table 6 presents the means and standard deviations for the eighteen attitude statements. State- ment 10 was selected for use as the topic the public figure wou1d argue against. This statement was "Federal spending ought to be curtailed in order to bring the national budget into balance." 49 TABLE 3 RESULTS (FAMILIARITY AND IMPORTANCE) FOR PUBLIC FIGURE SELECTION PRETEST Mean familiarity and importance ratings of pretest public figures (n=54).1 Public Figure Familiarity Importance Y Y Bella Absug 4.39 3.34 Birch Bayh 3.95 3.75 Edmond Brown 4.49 3.14 Bob Carr 3.61 3.12 Jimmy Carter 4.21 2.96 Shirley Chisolm 3.40 2.88 Gerald Ford 1.56 1.37 Hubert Humphrey 2.91 2.85 Henry Jackson 4.29 3.43 Edward Kennedy 3.12 2.62 Henry Kissinger 2.31 1.72 George McGovern 2.79 3.15 Edmond Muskie 3.61 3.06 Ronald Reagan 2.57 2.64 Elliott Richardson 4.53 3.60 Nelson Rockefeller 2.31 2.33 Sargent Shriver 3.22 3.58 Morris Uda11 4.46 2.95 George Wallace 2.36 2.53 1 Lower ratings represent greater fami1iarity and greater importance ‘ /‘\l 1 l . . \ i 50 TABLE 4 RESULTS (INFORMATIONAL SOURCES) FOR PUBLIC FIGURE SELECTION PRETEST Mean number of times 55 have had contact with the public figures by (a) media sources and by (b) interpersonal sources (n=54). Interpersonal Public Figure Media Infbrmation Infbrmation Bella Abzug 0.18 0.06 Birch Bayh 0.85 0.06 Edmond Brown 0.28 0.14 Bob Carr 0.49 0.00 Jimmy Carter 0.12 0.13 Shirley Chisolm 0.28 0.00 Gerald Ford 9.41 2.64 Hubert Humphrey 2.82 0.35 Henry Jackson 1.00 0.00 Edward Kennedy 1.62 0.58 Henry Kissinger 4.76 1.79 George McGovern 0.55 0.08 Edmond Muskie 0.12 0.07 Ronald Reagan 2.02 0.64 Elliott Richardson 0.00 0.08 Nelson Rockefe11er 3.21 0.35 Sargent Shriver 0.42 0.25 Morris Uda11 0.85 0.07 George Wallace 1.68 0.48 * The higher the number, the more often the 55 have heard about the public figure. 51 TABLE 5 RESULTS (IDENTIFICATION OF OCCUPATIONS) FOR PUBLIC FIGURE SELECTION PRETEST Occupations 55 reported for public figures and the frequency by which occupations were reported (n=54). 1 Public Figure Response Freguency Bella Abzug Don't know 21 N.Y. Congressperson 12 Birch Bayh Don't know 25 Senator (Indiana) 16 Edmond Brown Don't know 23 Governor of California 4 Bob Carr Don't know 15 CongreSSperson 16 Councilperson 6 Representative 4 Jimmy Carter Don't know 17 Governor of Georgia 6 Shirley Chiso1m Don't know 13 New Hampshire Congressperson 9 Gera1d Ford President 40 Hubert Humphrey Don't know 10 Retired Vice-President 10 Senator 7 Henry Jackson Don't Know 26 Senator 7 Edward Kennedy Senator (Massachusetts) 28 Don't know 12 Henry Kissinger Secretary of State 22 Statesperson 13 George McGovern N. 0. Senator 12 Ran for President 22 Don't Know 5 Edmond Muskie Senator (Maine) 21 Don't know 16 TABLE 5 (continued) Public Figure Ronald Reagan E11iott Richardson Nelson Rockefe11er Sargent Shriver Morris Uda11 George Wallace 1 52 Resgonse Governor (California) Senator Don't know Don't know Secretary of HEW Cabinet Vice-President Don't know Don't know Kennedy's Brother-in-Law Don't know Senator Secretary of Interior Governor of Alabama Presidential hopeful Don't know Only the more frequent responses are listed. Frequency ‘K‘Js "Wu-u. _ Means and standard deviations for 18 attitude statements (n=33). Attitude Statement 53 TABLE 6 RESULTS OF TOPIC SELECTION PRETEST .1... 1. It is the obligation of the Federal government to enable all American Citizens access to decent health care through a socialized medicine program ..... 4.03 2. Congress should drastically cut back on the inflated Pentagon budget .............. 4.12 . School busing to achieve racial balance in schools will worsen racial relations--not improve on them.. 3.85 . With the rapid development and production of nuclear arms, support for detente is crucial and imperative ..................... 3.35 5. Legislation should be enacted immediately to curtail the oil companies' gigantic profits at the consumer's expense ................. 3.18 6. The United States should withdraw all su ort, both arms and financial, from the conflTEt in ngola. . . 4.94 7. The time is long overdue to grant an unconditional amnesty to American men still in exile in foreign countries because of their stand against the Vietnam war ..................... 4. 47 8. Education is valuable for its own sake, even if it doesn't prepare you fbr a job ............ 4.05 9. The sale and use of marijuana should be legalized. . 4.97 10. Federal Spending ought to be curtailed in order to bring the national budget into balance ....... 4.00 11. The Federal government should take over ownership an and operation of all railroads in the United States. 7.35 Ss' Attitude s.d. 2.75 2.23 3.05 1.89 2.44 2.94 3.16 3.07 3.79 2.00 2.23 I C. \ O 54 TABLE 6 (continued) Attitude Statement 12. 13. 14. 15. 16. 17. 18. Any level of unemployment can be tolerated for a few years if it will help to end inflation . . . . The government should Spend as much money as necessary in order to start Federal programs to provide jobs for the unemployed who want to work . Candidates who run for public office should agree when they do so to surrender their right to privacy and expose all of their affairs to public scrutiny ...................... Government intelligence agencies should never en- gage in activities either at home or abroad which would be illegal if a private citizen did them . . No matter what the risks to national security may be, Congress and the public must be fully in- formed of the activities of all government agencies, including the C.I.A ................ The long range effects of pollution could be so devastating that any measures, no matter how costly in economic terms, are justified if they will help to stop industrial pollution ....... Air pollution in cities like Los Angeles and New York is so harmful to the health of the residents of those cities that they should outlaw all auto— mobile traffic ................... Ss' Attitude 'X 8.38 5.00 8.38 3.73 7.21 3.85 8.32 s.d. 2.36 3.02 2.75 2.83 2.98 3.28 3.28 The lower the mean, the more the 55 agree with the attitude expressed with the statement. 55 Message Design Pretest Having selected a target public figure, two familiar public figures and twelve unambiguous traits (end points of six credibility-related attributes) through the procedures discussed above, it was necessary to conduct a pretest to locate these concepts, plus the concept "Ideal Credible Source," in a multidimensional space. Since it may be the case that an abstract concept such as the "Ideal Credible Source" would be ambiguous to the §s, a brief description was presented in the cover letter: "Imagine that the "Ideal Credible Source" is an individual whoqyou LIKE TO HEAR SPEAK, whose_9pinions YOU TRUST and WHOSE ADVICE YOU WOULD LISTEN IQ," The attempt in this description was to present a description ab- stract enough to allow the §s to infer from it what they believe to be a proper characterization of the construct. It was felt permissable to in- clude the word "trust" since it was not included in the questionnaire. The other two characterizations--"1ike to listen to“ and "whose advice you would listen to"--are abstract enough to allow for the Ss' own inter- pretation. For example, subsumed under each of these descriptions, one may infer trustworthiness, competency, safety, attraction and/or dynamism, etc. On January 22, 1976 fifty-four §s completed a questionnaire which presented the criterion pair and asked the Ss to make ratio judgments of the relative dissimilarities of each possible non-redundant pair of these sixteen concepts. A metric multidimensional representation of the config- uration of these concepts was then generated, utilizing available multi- dimensional software (e.g., Woelfel's "Galileo" program, see Woelfel, 1974). In order to construct messages that would move the target public figure directly towards and directly away from the concept "Ideal Credible D /(\ 56 Source" the following procedure was employed. First, Rm was defined as the vector of each of the n concepts and R1 is defined as a message strat- egy based on vector addition of several of the vectors in the space. Procedurally, R0 (vector representing the point of the target pub1ic fig- ure) is set equal to zero. Functionally, this locates the target public figure at the centroid of the space. From this point, the vector extend- ing from the centroid to the desired point ("Ideal Credible Source") is identified. This can be referred to as the target-ideal point vector (Rm). Next, the angles between all other vectors and RI“ are calculated. Finally, resultant vectors of all pairs of concepts are calculated by a vector addition procedure. All single vector solutions, two pair vec- tor solutions, three pair vector solutions and fOur pair vector solu- tions were analyzed. Of all these resultants, the resultant vector whose angle with Rm is minimal will provide the concepts which, when utilized in the credibility induction messages, provide the maximally efficient means for moving the target public figure in the desired dir- ection. (For detailed elaboration of this procedure, see Woelfel gt_gl,, 1976). The solution utilized in the present study is presented graphically in Figure 5. The solution reveals that a message which described the target public figure as similar to Humphrey, just, competent and experi- enced would move the target public figure directly towards the desired point. The correlation between Rm and Ri (resultant vector) was 1.00. The angle was .99. Further, the solution also posits that if full effects of the messages were obtained, the target public figure would move beyond the desired point by 136.69 units. However, it is doubtful 57 44 mafia: mm.wman mpfias :fi.mm u mpfiqs m>.m: muwcs mo.mb mpwq: mo.m: muwnd bo.Ho m mo npmaea fimefl nemeoa npmema epmeea npmeea mm.o u e a a m m 0 oo.H u E -H g :OOQSOm manflfimfio HmmUH: zcooaowpoaxo: u En: u Mad : Pmd m: (03‘ zpcopomsoo: u hopzassm phenom u sham nopfim u C) r4 cw means when: .cowusHOm pouoo> mammlmsom mo cowumaprHHH .m madman 58 whether full effects could be obtained. What is important, and shall be tested (see Chapter III), is that the message moves the public fig— ure along this predicted vector (R1). Hence, the above fbur concepts were employed in the positive credibility induction message and the concepts unjust, inexperienced, incompetent and dissimilar to Humphrey were used in the negative cred- ibility induction message. In constructing the messages, a thesaurus (Roget's College Thesaurus, 1962) was used as an aid in selecting ad- jectives typically considered as similar in meaning to the selected vector-concepts. Alternatively, synonyms could have been pretested. However, in each message the selected vector-concept adjective was spec- ified with adjectives obtained from the thesaurus so, presumably, no additional attribute contaminated the results. Further, adverbs were selected that could apply to both positive and negative adjectives, sen- tence structure was standardized in both messages, and the placing of concepts at equivalent locations in the messages helped to make the messages exact opposites of each other. These messages appear in the Appendix. Posttest (a) Materials Two questionnaires were used in the posttest phase of the study. The first questionnaire consisted of twelve attitude state- ments similar to the eighteen used in the pretest for topic selection. Each statement was followed by an eleven-point Likert type scale to mea- sure the Ss' level of agreement with each statement. End points on these scales were reversed on every other statement to eliminate possible re- Sponse bias effects. This questionnaire provided the pretest measure of 59 agreement with the topic selected for manipulation; balancing the Fed- eral budget. There were four forms of the second questionnaire. The first form of the questionnaire included a cover letter, the positive induction message, the persuasive message attributed to the target public figure, several attitude items (to provide posttest measure), the MOS paired comparison questionnaire and semantic differential scales for the six attributes. In order to assess attribute salience, two unidimensional scales were included in the section of the questionnaire that included other unidimensional scales. These scales were seven-point scales of relevance and importance. The end points of the scales were periodic- ally reversed to eliminate possible response bias effects. The second form of the questionnaire included the negative induc- tion message instead of the positive induction message. The third form included only the persuasive speech as the manipulation. The fourth fbrm included neither induction message nor the persuasive Speech. The latter group served as the control group. In a Similar study to the one presented here, conducted in the Fall of 1975, an interview with the §§ revealed that many of them questioned the credibility of the source of the credibility induction message. Since this could be problematic in the present study, particularly in the negative credibility induction message, the credibility inductions were attributed to a hypothetical citizens' committee, briefly described to characterize it as non-partisan: The following description of Birch Ba h is taken from a pamphlet written and circulat53_by .onpartisan Citizens' Committee for the Promotion of Infbrmed Voting." This group is composed largely of indepen- dents and is active primarily in Minneapolis and In- dianapolis. The group has analyzed a large number 1 C {1 60 of nationally known political figures in terms of voting record, interests and qualifications. The following is what the group had to say about Birch Bayh: (b) Procedures On February 9, 10 and 12, 1976, 184 SS participated in the final phase of the study. During each of the three test ses- sions, §s first received the attitude questionnaire. When §s finished filling out the attitude questionnaire, they returned the questionnaire and the four forms of the second questionnaire were distributed in ran- dom order--thus SS were randomly assigned to four treatment groups within each of the test sessions. When §s completed the second ques- tionnaire, they were provided with a written description of the aims of the study. This description also served as the debriefing. (c) Randomization Check To check the random assignment of §§ to posttest groups a one-way ANOVA was computed on the pre-persuasion atti- tude scores of the four posttest groups. The obtained F-ratio was .72 (not significant) with means of 4.65, 5.37, 4.70 and 4.82 for the nega- tive, control, Speech only and positive groups, respectively. Thus, §s were adequately randomized into posttest groups on the basis of atti- tude scores on the topic which served as the manipulation check on the credibility manipulation. CHAPTER III RESULTS The results are divided into five sections: (1) Comparability of the Spaces; (2) Attribute Salience; (3) Hypotheses; (4) Manipulation Check; and (5) Message Effectiveness. Sections (1) and (2) provide ex- plicit and stringent tests of the proposed model. Sections (3), (4) and (5) provide tests of the experimental manipulations of the present study. Comparability of the Spaces One of the key assumptions of the model is that the trait adjec- tives will be stably located in the multidimensional space. Further, it has also been argued that there are as many dimensions as there are re- liable dimensions. In this section, I shall discuss three procedures that were utilized in assessing the stability of the concepts. The first procedure provided an overall measure of fit between the locations of the stable concepts in each of the groups with their loca- tions in an aggregate. This aggregate was created by aggregating togeth- er the distance estimates in all four experimental groups. The ration- ale behind the procedure is that the distances between stable concepts in each of the four posttest groups should approximate a cultural-level configuration. This procedure assesses the extent to which locations of these concepts in each posttest group tend to conform to such an aggre- gate. 61 62 The second procedure employed to assess stability of stable con- cepts was a cross-group correlation procedure. This procedure provided the correlations and angles between a concept and itself in a split-half of the control group, and between the control group and each of the three experimental posttest groups. This provided information as to each con- cept's stability. The third procedure utilized to assess comparability of spaces was to compute factor correlations between each of the sixteen factors in the control group and each of the sixteen factors in the three experi- mental posttest groups. This provided an assessment of the reliability or stability of each factor. Before these three procedures were employed, some prior data ad- justment procedures were utilized. Therefore, in this section of the paper- I shall discuss, in order: (a) Data Adjustment Procedures; (b) Overall Assessment of Fit; (c) Cross-Group Correlations; (d) Factor Cor- relations. (a) Procedures Utilized to Adjust Data As earlier research indicated (Barnett, Serota and Taylor, 1974), a multidimensional space can essentially “shrink" either over time or due to an experimental manipulation. In the present study, the shrink- ing phenomenon did occur and the means for all non-zero cells varied from group to group, with the control group having a “larger" space. The average distance reported by SS in the negative induction condition was 63.657, by SS in the positive induction condition, 67.911, by SS in the speech only condition, 69.556, and by SS in the control group, 87.846. This shrinking phenomenon affected all 120 distance estimates, including the distance estimate for the criterion pair. Since any 63 statistical analysis computed across groups would be biased due to this shrinking, all spaces were adjusted by an additive constant. For each space, an additive constant was computed by subtracting the obtained mean distance for the criterion pair from 100. The remainder was added to each of the 120 distance estimates in the respective condition. The additive constants used were 30.244, 12.154, 30.444 and 29.099 for the negative induction, control, speech only and positive induction groups, respectively. A second adjustment of the data was made before statistical anal- yses were conducted. In the metric MDS procedure, estimates of dis- tances become more accurate and reliable as the number of respondents increases. However, since data in the present study were collected from rather small samples, any reported distance that is extremely high, or is extremely low, will have a marked influence on the mean. For exam- ple, if, for 44 Ss, the mean of a particular distance estimate is 79.025, the inclusion of a value of 400 by the 45th §_increases the mean to 86.158. This could be extremely problematic in small samples because highly discrepant scores distort the true mean and unduly increase var- iance. A standard procedure employed when distributions have long, straggling tails is to trim the means - thus reducing the influence of highly discrepant scores (see Mosteller and Tukey, 1969). A means trim- ming procedure was executed in the present analysis by obtaining the means and standard deviations and eliminating all observations that are two standard deviations above and below the mean for each of 120 pairs of concepts. For this initial computation, distance estimates were left to vary from 0 to 999. After the means were trimmed, the obtained means 64 for the criterion pair were 96.93, 90.10, 93.49, and 101.53 for the negative, control, speech only and positive posttest groups. These values closely approximate the desired 100. Thus, the spaces were made comparable.* How does this means trimming operation affect the analysis of the interpoint distance between Birch Bayh and "Ideal Credible Source"? Table 7 presents the means and standard deviations for this distance distribution for each group. Further, the table reports the maximum value and minimum value used in computing the ANOVA, as well as the num- ber of observations deleted because of the means trimming operation. A smaller range of values was permissible for the negative induction group, in comparison with the other groups, but this was because the standard deviation was much smaller in comparison with that obtained in the other three groups. Only a few observations were deleted from each group. (b) Overall Measure of Fit To test the stability of the stable concepts 'hi the multidimen- sional scale, the following procedure was utilized. First data from the four posttest groups were aggregated and a space generated which estim- ated the trait, Ideal Source configuration which each group's space Should approximate. Spaces were then generated from each group's data and rotated into a least-squares best fit of the theoretically expected stable points (i.e., the traits and the "Ideal Credible Source") with * In future research, it would be preferable, in small samples, to first trim means and then add the additive constant. The means for the criterion pair are comparable in this study, reversing the order of these procedures would ensure a value of 100 for the criterion pair distance. 65 TABLE 7 RESULTS OF MEANS TRIMMING PROCEDURE ON THE BIRCH BAYH-IDEAL CREDIBLE SOURCE DISTANCE DISTRIBUTION Means, standard deviations, minimal permissible values, maximum per- missible value and number of cases deleted because of means trimming procedure (Birch Bayh-Ideal Credible Source distance estimate). Number of 01d* New Standard Minimum Maximum Observations Mean Mean Deviation Value Value Deleted Negative 94.97 95.21 35.00 25.00 165.00 2 Control 88.66 83.02 50.00 0.00 188.00 2 Speech Only 90.56 86.02 45.00 0.00 180.00 3 Positive 77.31 68.65 50.00 0.00 177.00 2 * Represents the value of the means for each posttest group before the means trimming procedure was implemented. 66 the aggregated space (see Woelfel gt_gl., 1975). Correlations were then computed between the coordinates defining the location of each stable point in the aggregate space and the corresponding vector in each group space. These 52 (13 Stable Points x 4 Group Spaces) correlations were then converted to zfs, averaged, and the mean z_was reconverted to a correlation. This correlation measures the extent to which the confir- uration of stable points in each group space approximates the same pat- tern or shape regardless of its absolute size. There exists a certain amount of redundancy in this method since the raw data from each post- test group is used in creating the aggregate. This would tend to skew correlations upwards. To support the test for stability, this overall correlation must be extremely high. Table 8 provides correlations and z_scores between coordinates of each concept on the first nine dimensions (most of the dimensions repre- senting real distances) of the aggregate and each posttest group. These nine dimensions accounted for the following percentage of variance in the negative induction, control, Speech only and positive induction groups, respectively: 97.4, 98.4, 98.5, and 98.3. The mean 5 score was 2.84, which, when reconverted back to a correlation, yields a cor- relation of .993. Data presented in Table 8 does not include the means trimming operation. When input was restricted to a range of 0 to 180, the mean z_was 2.818--a correlation of approximately .993. (The metric multidimensional scaling program used in this study (see Woelfel, 1974) does not possess options for either multiple maximum values or multiple minimum values. The above parameters, a minimum of 0 and a maximum of 180, were used because it was felt that these parameters best approxim- ated the average minimum and maximum for all distance distributions.) 67 TABLE 8 CORRELATIONS AND Z_SCORES FOR STABLE CONCEPTS. Correlations and z_scores between a concept's coordinates in the aggre- gated data set and each of the posttest groups. Aggregate- Aggregate- Aggregate- Aggregate- Stable Negative Control Speech Positive Concepts Condition Condition Condition Condition .9923 .9820 .9965 9912 Competent 2.7790 2.3510 3.1730 2.7110 .9994 .9953 .9941 .9872 Inexperienced 4.0560 3.0250 2.9110 2.5220 .9949 .9949 .9676 .9263 Repulsive 2.9850 2.9850 2.0530 1.6320 .9956 .9901 .9907 .9806 Unintelligent 3.0580 2.6520 2.6830 2.3130 .9965 .9982 .9984 .9984 Just 3.1730 3.5060 3.5650 3.5650 .9987 .9955 .9887 .9719 Reliable 3.6690 3.0470 2.5850 2.1250 .9988 .9946 .9907 .9841 Unreliable 3.6690 2.9560 2.6830 2.4130 .9950 .9988 .9988 .9840 Unjust .9950 .9988 .9988 .9840 .99814 .9955 .9937 .9443 Intelligent 3.4900 3.0470 2.8790 1.7760 Ideal Credible .9847 .9904 .9902 .9893 Source 2.4330 2.6670 2.6570 2.6120 .9984 .9890 .9709 .9391 Attractive 3.5650 2.5940 2.1080 1.7300 .9947 .9922 .9956 .9880 Incompetent 2.9650 2.7710 3.0580 2.5550 .9953 .9946 .9939 .9789 Experienced 3.0250 2.9560 2.8950 2.2700 Mean 2 = 2.84 standard deviation = .56 r = .993 68 To assess the relative stability of the thirteen theoretically stable points in both real and imaginary dimensions, the procedure was used in correlating the coordinates of each concept across all sixteen dimensions between the aggregate and each posttest group. The obtained mean z_was 2.584--a correlation of .988. Thus, the concepts which we have argued as theoretically stable appear to be highly stable. (c) Cross-Group Correlations An alternative and more stringent procedure for assessing the sta- bility of the stable concepts is to compare each of the experimental groups with the control group. Cross-group correlations were computed by treating the control group configuration as the first point in time and by performing separate rotations for each of the experimental group configurations to it with the thirteen theoretically stable concepts specified as stable concepts (see Woelfel gt_al,, 1975). Correlations were then computed between the concept's location in the control group configurations and its location in each of the experimental group con- figurations. These cross-group correlations are presented in Table 9 for a split-half of the control group, the speech only, negative induc- tion and positive induction group comparisons, respectively. Befbre presenting these correlations, it would be worthwhile to discuss briefly why these correlations are important. Cross-group com- parisons provide infbrmation concerning how well a concept correlates with itself between groups (the angles between the vectors that repre- sent the concept in the two groups). When the angle is greater than zero there is a departure from stability: an angle that is greater than zero represents a change in the concept's “meaning“ relative to 69 other concepts in the configuration. Of course, the stable concepts cen- troid is derived by finding the least-squares best fit between concepts identified as stable (see Woelfel gt_al:, 1975). If some stable concepts have expanded or contracted, this would influence the location of the stable concepts centroid to some small degree. Therefore, some small angles will be obtained artifactually. All angles in a cross-group com- parison must be interpreted relative to other angles in the same cross- group comparison. Table 9 presents the results of the cross-group correlation proce- dure for a Split-half of the control group, control group-Speech only group, control group-positive induction group, control group-negative induction group comparisons. While the sample size in the split-half of the control group is small (each n=22), the stability correlations presented in Table 9 can be used as a baseline for comparison with the stability correlations presented in the remaining three columns of Table 9; where correlations are affected by both unreliability and by message(s) used as experimental manipulations. Correlations presented in the first column of Table 9 provide evidence for the stability of the stable con- cepts, with the exception of "attractive," "just“ and "repulsive" (correl- ations of .65, .72 and .76, angles of 49.78, 44.09 and 40.28 degrees, re- Spectively). Nonetheless, the remaining ten correlations are remarkably high considering that the samples sizes are so small. The second column in Table 9 presents the results of the cross-group correlation procedure for the control group-speech only group comparison. First, it should be noted that angles cannot be computed for ten of the thirteen stable concepts (correlations are greater than one). This occurs for the following reason. When a concept loads higher in the 70 TABLE 9 CROSS-GROUP CORRELATIONS BETWEEN EACH STABLE CONCEPT Correlations and angles between each concept for (a) split-half of con- trol group, (b) Control Group-Speech Only Group, (c) Control Group- Positive Induction Group, and (d) Control Group-Negative Induction Group.* Split-half Control- Control- Control- Concept Control Speech Positive Negative Competent (Ol) .99 1.08 1.12 .95 7.09 **** **** 18.21 Inexperienced (02) .99 1.02 .93 1.00 7.80 **** 21.58 **** Repulsive (04) .76 .96 .87 .81 40.28 15.79 29.27 36.09 Unintelligent (05) .92 1.19 .77 .80 22.61 **** 39.79 37.20 Just (06) .72 1.19 .76 1.01 44.09 **** 40.81 **** Reliable (07) .91 1.02 1.11 .98 24.39 **** **** 10.56 Unre1iab1e (08) .88 .99 1.04 .99 27.66 5.06 **** 8.72 Unjust (10) .80 1.12 .88 .94 36.75 **** 28.30 20.14 Intelligent (11) .97 1.42 .89 .70 12.51 **** 27.13 45.49 I. C. S. (12) .86 1.09 1.01 .96 30.76 **** **** 16.35 Attractive (I3) .65 .97 .89 .91 49.78 14.31 26.36 23.99 Incompetent (14) .99 1.08 1.05 .98 8.70 **** 17.20 9.99 Experienced (15) .81 1.07 .95 1.01 35.55 **** 17.20 **** * The first number is the correlation; the second number is the angle. Asterisks represent "negative“ angles. 71 imaginary part of the multidimensional space in the experimental group than in the control group, the scalar product lengths are reduced, thus increasing the correlation. Correlations slightly greater than 1.00 will be interpreted as representing stability. A correlation that is moderately or substantially larger than 1.00 will be interpreted as rep- resenting a departure from stability-~specifically, instability in the imaginary part of the space. 0f the concepts listed in the second col- umn of Table 9, only "intelligent" appears to be unstable (correlation = 1.42). The third column in Table 9 presents the results of the cross-group correlation procedure for the control group-positive induction group comparison. Among the stable concepts, "just," and “unintelligent" rep- resent some departure from stability (angles of 40.81 and 39.79). How- ever, these angles represent correlations that are still relatively high (.76 and .77, respectively) and can hardly be referred to as significant departures from stability. The fourth column in Table 9 presents the results of the cross-group correlation procedure for the control group- negative induction group configuration. Among the stable concepts, "in- telligent," "unintelligent," and "repulsive" represent some departure from stability (angles of 45.49, 37.20, and 36.09). These angles repre- sent stability correlations of .70, .80 and .81, and can hardly be re- ferred to as major departures from stability. In general, it appears that the results of the cross-group correla- tion procedure provides strong evidence for the stability of the concepts. A considerable number of correlations were in the .9's across the compar- isons and most correlations were between .9 and 1.1. While the poorest correlation was indeed small (.65), it was obtained in the split-half of 72 the control group and the concept ("attractive") stabilized in compari- sons with larger samples (.97, .89 and .91) in the speech only, positive and negative group comparisons. However, two concepts appear to depart, to some degree, from stability in more than one group: "just" correlated .72 and .76 in the split-half comparison and the control group-positive induction group comparison; and "intelligent" correlated 1.42 and .70 in the control group-speech only group, and control group-negative induction group comparison. Nonetheless, the severity by which these concepts de- parted from stability in two out of the four comparisons does not warrant rejecting them from the stable concepts rotation. (d) Factor Correlations I have presented evidence for the stability of the stable concepts across groups as well as between an aggregate and each group separately. In this section, I would like to assess the stability of each factor structure. Procedurally, the calculation of factor correlations utilizes the sets of coordinates from each control group-posttest group stable concepts rotation and computes the correlation between the set of concept loadings in each of the posttest groups. The procedure was also used fbr a Split-half of the control group. For the obvious reasons that the loadings of the three person-concepts (Bayh, Humphrey and McGovern) change in all conditions, only the thirteen stable concepts were utilized in the calculation of these correlations. Factor correlations calculated using a Split-half of the control group provide infbrmation about the factor structure stability without effects of the messages used as manip- ulations. Factor structure stability correlations calculated between the control group and each posttest group indicates how stable and robust 73 the factor structure is even when the experimental messages utilized several of the concepts. Table 10 presents the results of the factor correlation procedure. Although the Split-half of the control group resulted in small samples (each n=22), the factor correlations fer the first three factors (67.46 percent of the variance) and the last three factors (15.00 percent of the 21.71 percent of imaginary variance) provide strong evidence for the stability of the factor structure for a substantial part of the multi- dimensional Space. For a larger sample, where the concepts may have been influenced only by the persuasive speech (see second column of Table 10), there is evidence that the structure of the first five real factors are stable. These five factors account for 85.12 percent of real variance. Factor correlations computed between the control group and the nega- tive induction group indicate that the first Six factors are stable (accounting for 90.83 percent of the real variance). Unfortunately, less factor structure stability was obtained for the control group-positive induction group comparison. Only the first three factors are stable, accounting for 67.46 percent of the real variance. The loss in factor structure stability is presumably due to the effects of the persuasive Speech and the positive induction message. Apparently, attributing "Humphrey," “just," "competent" and "experienced" to Birch Bayh in the positive induction message had a much stronger influence on the stability of the factor structure of the concepts than did attributing "dissimilar to Humphrey," “unjust," "incompetent" and "inexperienced" to Birch Bayh in the negative induction message. In general, it can be concluded that between three to five of the real factors are stable, and that the last three (imaginary) factors are “l Iva—.mp— -... 74 TABLE 10 RESULTS OF FACTOR CORRELATION PROCEDURE Correlations between loadings of stable concepts in each of 16 factors for (a) split-half of control group, and fOr control-group with loadings in (b) speech only group; (c) positive induction group; and (d) negative induction group. Percent Control Control- Control- Control- Variance Split- Speech Positive Negative (Control Factors Half Only Induction Induction Group) 1 .99 .99 .98 .98 42.94 2 .96 .98 .97 .97 13.62 3 .94 .93 .89 .83 10.90 .66 .92 .64 .86 9.61 5 .45 .93 .58 .87 8.05 6 .45 .20 -.OO .95 5.71 7 -.11 .75 .14 -.54 4.38 8 .01 .97 .80 .90 3.06 9 -.64 .91 -.39 .23 1.57 10 .70 .79 .84 .83 .153 11 -.15 .07 -.OO -.66 .OO 12 -.22 -.72 .89 .64 -2.15 13 .31 .14 .76 .60 -2.56 14 .77 .75 .77 .86 -4.43 15 .93 .77 .80 .66 -5.21 16 .96 .48 .98 .99 -6.36 75 stable. Perhaps the best indication of the real factor structure stabil- ity is represented by the factor correlations obtained in the control group-speech only group comparison. In this comparison, the sample sizes are larger than in the Split-half of the control group, and the locations of the concepts are influenced only by the speech. Thus, it can be con- cluded that the first five factors are fairly stable. The factor struc- ture stability correlations for the control group-negative induction group comparison supports this conclusion. Unfortunately, the weak correla- tions for the fourth and fifth factors in the control group-positive in- duction group comparison (correlations of .64 and .58) argues against this conclusion. However, in comparison with the control group-speech only group, and the control group-negative induction group correlations, it is safe to say that the lack of stability of these two factors is not due to unreliability of measurement, but,rather, to the effects of the persuasive speech and positive induction message. Finally, it appears that the factor structure correlations fOr the last three (imaginary) factors are fairly robust. It is not surprising to obtain factor structure instability fbr factors 6 to 13. These factors account for only 14.87 percent of the real variance and only 4.71 of the 20.71 percent of imaginary variance. Eigenvalues for these factors are considerably smaller in comparison with the eigenvalues for stable factors. For example, the eigenvalue for factor eleven is -.039 and this factor accounts for -.000 percent of variance (negative induction group). Factor eleven is a null vector, and represents only machine and programming rounding error. The eigenvalue for the largest unstable factor (factor 6) is less than 4000.0 and accounts for less than six percent of the variance. (See Tables 18-20 for III an“... ..-1 1.1., 76 eigenvalues for all factors). In comparison with the stable factors, the unstable factors account for a trivial amount of infbrmation about the interpoint distance between concepts. Attribute Salience In proposing the model, it was argued that the smaller the pro- jected line from a concept to the line segment anchored at both ends by bipolar adjectives, the more salient the attribute is in the Ss' per- ception of that concept. A computer program was written to find the length of this projected line from each of the four concepts ("Ideal Credible Source," "Birch Bayh," "Hubert Humphrey," and "George McGovern"). The length of each projected line was subtracted from 100 so that the higher the value, the higher the salience score. This measure of sali- ence was compared with the unidimensional ratings of relevance and impor- tance. Tables 11, 12, 13 and 14 provide the multidimensional model's de- rived salience measure and the mean ratings on the unidimensional scales of relevance and importance for five groups of §§ (pretest and fOur post- test groups) for all four of the concepts on the six attribute line- segments. Two correlations were computed across the five groups of £5 for each concept summed across attributes: a correlation between sali- ence scores and the mean relevance rating and a correlation between sali- ence scores and mean importance ratings. Further, an overall correlation across groups and across the four concepts was computed for both salience- relevance and salience-importance pairs of ratings. Thus, a total of ten correlations were computed. 1.. .111.|11.F1.I 77 TABLE 11 SALIENCE SCORES FOR BIRCH BAYH Salience scores, mean ratings on relevance scale and mean ratings on importance scale.* Pretest Control Sgfi$5h Positive Negative n=54 n=47 n=45 n=45 n=47 Competent salience (1.1112) 75.70 50.71 61.60 55.55 01) Incompetent relevance 4.39 4.55 5.04 5.29 5.11 (14) importance 3.76 3.84 4.38 4.81 4.38 Inexperienced salience 71.22 87.80 56.73 60.23 62.49 (02) Experienced relevance 3.74 3.78 4.28 4.49 4.73 (15) importance 4.35 4.88 5.55 5.11 5.06 Repulsive salience 49.51 57.55 28.31 36.97 38.56 (04 Attractive relevance 3.52 3.62 3.70 3.40 3.57 (13) importance 3.78 2.62 2.81 2.51 2.47 Unintelligent salience 64.39 71.40 49.64 59.25 46.20 (05 Intelligent * relevance 3°17 3-80 4.75 5.16 4.41 (11) importance 4.07 3.96 4.66 5.33 4.68 Just ) salience 61.40 70.06 51.50 56.15 46.81 06 Unjust relevance 4.50 4.44 5.28 5.09 5.13 (10) importance 3.96 3.67 4.53 4.89 4.36 Reliable salience 76.13 75.47 50.96 49.79 44.01 07) Unreliable relevance 4.48 4.96 5.51 5.18 5.08 (08) importance 3.88 4.11 4.64 5.47 4.49 * Higher ratings or scores represent higher values of importance, rele- vance and salience. Numbers in salience score columns that are inside parentheses were salience scores which could not be computed. These observations were not used in calculating correlations. Numbers in parentheses represent correlations greater than one. The computer program that provides attribute salience scores does not, at this time, provide the shortest path to one of the bipolar adjectives. I 4/ 78 TABLE 12 SALIENCE SCORES FOR HUBERT HUMPHREY Salience scores, mean ratings on relevance scale and mean ratings on importance scale.* Pretest Control Sgfi$5h Positive Negative n=54 n=47 n=45 n=45 n=47 Competent salience 68.20 58.81 46.98 56.96 54.37 Inégm)etent relevance 4.28 5.31 5.11 5.98 4.59 (14 importance 4.54 4.64 4.55 5.13 4.70 Inexperienced salience (1.014) 84.00 81.22 70.47 75.78 Exégiienced relevance 4°434 3-91 4.13 4.80 4.34 (15) importance 4.811 5.02 5.40 5.36 5.40 Repulsive salience 42.39 35.52 22.94 31.84 25.67 Attggltive relevance 3.79 3.40 3.32 3.33 3.57 (13) importance 2.72 2.82 2.43 2.67 2.53 Unintelligent salience 91.08 74.12 51.73 54.43 49.73 In£gfiiigent relevance 4.770 4.71 4.68 5.24 4.94 (11) importance 4.878 4.71 4.72 5.33 5.06 Just salience 59.50 64.01 54.08 53.48 43.37 Unjggt relevance 4.88 4.62 5.02 5.13 4.87 (10) importance 4.83 4.51 4.66 4.82 4.89 Reliable salience 76.13 76.61 51.97 41.43 40.08 Unigiiable relevance 4.900 5.07 4.98 5.67 5.40 (08) importance 4.690 4.71 4.98 5.24 5.17 * Higher ratings or scores represent higher values of importance, rele- vance and salience. Numbers in salience score columns that are inside parentheses were salience scores which could not be computed. These observations were not used in calculating correlations. Numbers in parentheses represent correlations greater than one. The computer program that provides attribute salience scores does not, at this time, provide the shortest path to one of the bipolar adjectives. 79 TABLE 13 SALIENCE SCORES FOR IDEAL CREDIBLE SOURCE Salience scores, mean ratings on relevance scale and mean ratings on importance scale.* Pretest Control 532$§h Positive Negative n=54 n=47 n=45 n=45 n=47 Competent salience (1.1817) 56.99 74.02 63.51 70.33 01) Incompetent relevance 5.56 6.11 6.45 5.89 6.02. (14) importance 5.49 5.78 5.49 5.36 5.06 Inexperienced salience 70.36 67.01 67.78 69.89 70.31 02 Experienced relevance 5.02 5.09 5.23 4.84 4.74 (15) importance 5.48 6.29 6.30 5.73 5.62 Repulsive salience 63.06 36.04 26.92 31.94 29.11 (04) Attractive relevance 3.62 4.18 3.87 3.67 3.40 (13) importance 3.43 3.58 3.34 2.96 2.76 Unintelligent salience 79.41 63.25 57.33 59.77 56.72 05 Intelligent relevance 5.23 5.38 5.08 5.40 5.00 (11) importance 5.56 5.67 5.34 5.47 5.19 Just ) salience (1.446) 62.49 57.33 60.75 62.52 06 Unjust relevance 5.38 6.13 5.70 5.62 5.64 (10) importance 5.49 5.59 5.25 5.16 5.00 Reziaple salience 70.26 70.46 66.38 61.53 65.05 07 Unreliable relevance 5.56 6.49 6.21 5.98 6.23 (08) importance 5.68 5.73 5.57 5.49 5.38 * Higher ratings or scores represent higher values of importance, rele- vance and salience. Numbers in salience score columns that are inside parentheses were salience scores which could not be computed. These observations were not used in calculating correlations. Numbers in parentheses represent correlations greater than one. The computer program that provides attribute salience scores does not, at this time, provide the shortest path to one of the bipolar adjectives. ll. _./.\._ Div .t 80 TABLE 14 SALIENCE SCORES FOR GEORGE MCGOVERN Salience scores, mean ratings on relevance scale and mean ratings on importance scale.* Pretest Control Sgfieph Positive Negative n=54 n=47 n=45 n=45 n=47 Competent salience 66.35 63.00 53.53 58.06 50.30 (01) Incompetent relevance 5.090 5.55 5.85 5.84 5.36 (14) importance 4.897 5.09 5.29 5.33 5.08 Inpxperienced salience (1.100) 68.97 57.91 60.81 62.05 02 Experienced relevance 4.36 4.53 4.53 4.62 4.36 (15) importance 4.77 5.51 5.43 5.51 5.53 Repulsive salience 58.12 48.24 28.76 39.54 32.06 04 Attractive re‘evance 3.62 3.71 3.53 3.31 3.51 (13) importance 2.96 2.64 2.53 2.24 2.57 Unintelligent salience 64.73 59.46 50.53 53.70 49.67 (05 Intelligent relevance 4.72 4.58 5.04 4.91 4.81 (11) importance 5.09 4.84 5.02 5.20 4.87 Just ) salience (1.087) 79.42 57.19 49.81 44.04 (06 Unjust relevance 5.13 5.29 5.66 5.58 5.38 (10) importance 4.83 4.58 4.79 4.84 4.78 Reliaple salience 75.55 73.76 50.58 60.03 49.53 (07 Unreliable relevance 5.34 5.64 5.81 5.78 5.43 (08) importance 4.98 5.00 5.25 5.40 4.83 * Higher ratings or scores represent higher values of importance, rele- vance and salience. Numbers in salience score columns that are inside parentheses were salience scores which could not be computed. These observations were not used in calculating correlations. Numbers in parentheses represent correlations greater than one. The computer program that provides attribute salience scores does not, at this time, provide the shortest path to one of the bipolar adjectives. _ . ,\ '1 \, (x! .1 [I'm—- - 81 Table 15 presents these correlations, along with the means on each scale (summed across the attributes and groups of SS), standard devia- tions and number of usable observations.* In general, correlations be- tween salience scores and mean relevance ratings were lower than correl- ations between salience scores and mean importance ratings. The central reason for this seems to be that SS did not discriminate along the uni- dimensional relevance rating scale as much as they did on the importance rating scale. This is represented by consistently lower standard devia- tions on the relevance scale than on the importance scale (Table 15). This may be explained by pointing out that "relevance" may be a more ab- stract concept from the Ss' perspective. We shall focus on the relationship between salience scores and im- portance ratings. The relatively small correlation (.279) between the two types of scales for Birch Bayh is not surprising. Birch Bayh was only moderately familiar to the §§ and there was no correspondence be- tween the derived salience measure and the unidimensional ratings. The correlations between the two measures for more familiar concepts tend to support the salience measure. For George McGovern the two measures correlate .592. For Hubert Humphrey, the correlation is .627. Finally, for "Ideal Credible Source" the correlation is .825. Summing across all groups, attributes and concepts, the overall correlation is .578. This correlation is obviously reduced due to the inclusion of Birch Bayh ra- tings; nonetheless, it is reSpectably high enough to conclude that the multidimensional model's derived salience measure is tapping attribute salience. * See footnote in Tables 11, 12, 13 and 14 explaining unusable obser- vations. 82 TABLE 15 CORRELATIONS BETWEEN SALIENCE SCORES AND RATINGS 0N RELEVANCE AND IMPORTANCE SCALES Means and standard deviations on (a) salience, and (b) relevance or importance, number of observations and correlations. Birch Bayh Hubert Humphrey Ideal Credible Source George McGovern Salience (a) Relevance (b) u: >q 1 us >4 23709911 57. 13. .45 0. 4 29 -O 56. 17. .64 O. 4 29 O. 60. 13. 5. O. 28 O. 55 11 4 28 O. 45 O6 67 .026 44 25 679 252 37 19 28 860 663 .92 .59 .91 O. 77 442 Salience Importance (b) (a) 57 13 4 29 O. 56. 17. 4. O. 29 O 60. 13. 5. O 28 0. 55 11 4 O 28 O .45 .06 .24 0. 819 279 44 25 51 905 .627 37 19 12 .94 825 .92 .59 .64 .994 .592 1 | .lllllirv' 83 TABLE 15 (continued) Salience (a) Salience (a) Relevance (b) Importance (b) Overall Y; = 57.54 7; = 57.54 Sa = 14.06 E? = 14.06 Y5 = 4.81 Xb = 4.63 5n = 0.809 5b = 0.965 n = 114 n = 114 r = 0.338 r = 0.578 84 Let us briefly describe the attributes §§ perceived as salient in their conception of the "Ideal Credible Source" (Table 13). Means were computed for salience scores across the five groups of Ss for each attri- bute. Experienced-Inexperienced was the most salient attribute (69.070), followed by reliable-unreliable and competent-incompetent (66.74 and 66.21, respectively). The mean salience f0r intelligent-unintelligent was 63.29 and the mean salience for the attribute just-unjust was 60.77. The high mean salience scores indicate that among the six attribute line segments included in the multidimensional scope the five listed above were fairly salient in the Ss' conception or perceptions of the "Ideal Credible Source." The mean salience score for the attribute attractive- repulsive was only 37.41. This was not a salient attribute. Hypotheses To test the hypothesis that higher or lower credibility was system- atically induced, a one-way analysis of variance on the four posttest groups' distance estimates between the target and Ideal Source was com- puted. AS indicated in Table 16 (F=4.925, df=3/161, p<.01), the credi- bility inductions clearly altered the perceptions of Birch Bayh's credi- bility. The means for each group were 95.21, 83.025, 86.025, and 68.65, for the negative induction, control, speech only and positive induction groups, respectively. In addition, three contrasts were computed by the Scheffe post-hoc procedure. The first contrast compared the means between the control group and the negative induction group. The second contrast compared the means between the control group and the positive induction group. Neither of these contrasts were significant. The contrast between the I C I ‘1 1/ I 1111111111...." 85 TABLE 16 RESULTS OF ONE-WAY ANALYSIS OF VARIANCE Analysis of variance for Birch Bayh - Ideal Credible Source distance estimates. Source §§_ _d_f_ _1'_1_S_ f_ p Total 174,391.248 164 Between 14,657,626 3 4,885.875 4.925 .01 Within 159,733.622 161 992.134 86 means for the positive induction and negative induction groups was sig- nificant (Scheffe T = 3.768, Critical value = 2.793, p<.05). Thus, according to the F-test, procedures employed in the present study successfully altered the perceptions of Birch Bayh's credibility. Further, since Bayh moved 14.375 units closer to Ideal Credible Source due to the positive induction message and 12.185 units further away from Ideal Credible Source due to the negative induction message, it would appear that the positive induction message had slightly more impact. Yet, such a small difference cannot be cited as strong support for the notion of a bias in the Ss' processing of information such that they give a public figure in the negative induction message a benefit of a doubt. One should be cautioned against interpreting the findings of the F- test as final evidence. The F-test utilizes only one set of paired com- parison interpoint distances out of the 120 in the present study. It is only based on the interpoint distances between Bayh and the "Ideal Cred- ible Source," and does not take into consideration the direction Bayh moved due to the experimental manipulation. The directions Bayh moved will be assessed in the section entitled "Message Effectiveness." Manipulation Check The hypothesis which stipulated that the public figure who is closer to the "Ideal Credible Source" will stimulate more attitude change than a public figure who is more distant from the Ideal Source was assessed by means of a one-way analysis of variance across the attitude change scores in each of the four groups. Attitude change scores were computed by sub- tracting the posttest scores from pretest scores. As indicated in Table 17 (F = 4.169, df = 3/177, p<.01), the amount of obtained attitude change clearly varied across the four groups. The 87 TABLE 17 ANALYSIS OF VARIANCE FOR ATTITUDE CHANGE SCORES Source §§_ g:_ MS. 5. .p Total 180 1,390.69 Between 91.792 3 30.597 4.169 .01 Within 1,298.898 177 7.338 88 means for attitude change were 0.90, -0.64, -0.87 and -0.80 for the con- trol, negative, speech only and positive induction groups, respectively. Negative scores indicate change in advocated direction. Three contrasts were computed. One between the means for control group and negative induction group. The second contrast was between the means for the negative induction and the positive induction group. Neither of these contrasts were significant. The contrast between means for the positive induction group and the control group was significant (Scheffe T = 2.934, critical value = 2.793, p<.05). These results indicate that the persua- sive message was persuasive by itself. There was little difference be- tween the three experimental groups. The positive credibility induction evidently led to more attitude change than that obtained in the negative credibility induction, but this difference was not significant. Message Effectiveness In a previous section of the thesis ("Hypotheses") it was concluded that messages based on procedures described in Chapters I and II success- fully moved the manipulated public figure as predicted. The central criticism of this conclusion is that perhaps any positive or negative de- scription of Birch Bayh may have produced the obtained effects. We have hypothesized that the public figure would move along a specific resultant vector in the positive induction group and opposite to this specific re- sultant vector in the negative induction group (or, a resultant vector for the negative c0ncepts--dissimilar to Humphrey, unjust, incompetent and inexperienced), and this section of the thesis will assess the ex- tent to which Birch Bayh's movement in the space approximated the pre- dicted movement along the resultant vector. First, I shall briefly [C U 89 discuss the three dimensional visual representations (three dimensional plots) and how these representations illustrate the obtained motions. Second, I shall discuss how message effectiveness is derived and the ex- tent to which Bayh moved as predicted. Tables 18, 19, 20 and 21 provide the coordinates of the sixteen con- cepts in all sixteen dimensions for the control group, speech only group, positive induction group and negative induction group configurations, respectively. Figures 6, 7 and 8 present plots of the first three dimen- sions of each of the experimental posttest groups rotated to the control group with the thirteen stable concepts specified as stable (see Woelfel gp_pl,, 1975). Figure 6 presents the control group-speech only compari- son. Figures 7 and 8 present the control group-positive induction group and the control group-negative induction group comparisons, respectively. The first three dimensions account for the following percentage of var- iance in each group: 69.3, 64.8, 70.7 and 65.1, in the control, speech only, positive induction and negative induction groups, respectively. Concept identification numbers fer the concepts can be found by referring to Tables 18-21. The concept identification in each of the three figures is located by the concept's location in the control group confiruration. The second point on the line represents the concept's location in each of the posttest group configurations. Figure 6 illustrates the motions of Bayh (concept 3), Humphrey (9) and McGovern (16) that are due to the Speech. It would appear that Humphrey and McGovern move toward the negative part of the semantic space. (Indeed, McGovern's movement appears to be rather consistent across the three comparisons--Figures 6, 7 and 8). We shall return later to analyze Humphrey's movement; however, we are presently interested in how these K_ N \/ 90 mmm.mpm~ mvm.em nmo.m upm.mp www.mu Npm.o~1 mow.m oom.~ Nwo.mn www.mpn mmm.op Nm~.m mum.mu mmm.p_ mou.¢ mno.su Pum.-1 1m. PmF.mmmN mm~.o~u mm¢.m memaou omn._ NFF.N1 mum. emm.m- www.mmu N¢m.¢ Nmp.~F1 mmo.m _Nm.F1 w¢_.e— nmw.mm mnm.~1 N¢N.m M 0mm.—omm www.mm pmm.¢n Nnm.mpu mm~.n mnm.~u www.mu Num.w1 omm.mpu mcm.wp 05¢.up1 men.mp ppo.m mpm.¢ ~P¢.m~1 moe. omm.¢~ m Nom.mwme wmo.m uom.-1 mmo.mps mmm.e~ oo_.mu mmm.¢F mmm.Fm ¢NM.NN1 omu.o1 o¢~.op nmo.nma 5mm.mpu mmm.m1 ow~.~ mm~.m~ mmn.m— m me.¢-o omo. mam.- om¢.~u mme.e~ Nmm. oFm.mmu m¢N.mp -~.~_u moo.mu mop.pp mmm.mpu wep.~m mm¢.¢mu mop.p mmm.u~1 mmm.1 .41 omp.~m¢n o¢~.w mo~.op pmm.mm mn—.p¢ www.w moo.FN epm.pu P~¢.~n NNm.~m m¢¢.~ms omo.mu nmp.smu NFO.N~1 mm_.1 Npm.~m1 www.mmu m mmhwmpsamm exam cugwm coucmwcmaxmcm ucmumasou mmapm>cmmwm cgm>ouuz .mF umucmwsmaxm .m_ pneumaeoucu .ep o>wuumguu< .m_ .m .u .H .NP ucomequca .F— “name: .0? hugga53: 05 page r—NMQ'LOKONQ mummucou 91 mwo.m~mmu mwm.ou mmN.N~u owe.¢ mm¢.m Poo.~u FN¢.MN1 mmm.m sum.1 www.mm «mm.mm wom.PP oom.¢mn mm~.n mmm.~1 wom.m~1 mam.P mp www.mmvmu moo. oom.w1 map.mpn nm¢.F1 ¢u~.n mmo.N1 nmo.mm mwm.¢ FN~.FP1 mpw.m~1 mo~.mm owm. nmm.N1 ono.m Nmm.wu ¢m~.mpu mp mmm.P~om1 mum.¢ mpm.omu mom.¢ mmF.m1 emm.1 www.mm cup.m1 mam. omo.m mm¢.m mmm.1 umn.¢~ oom.m1 upn.v mom.¢mu mwm.m «F opm.ompmn omm.p oup.mpi mmm.NN mo¢.m mon.m nop.mu Pm¢.p omm.m mmm.npu www.mpu mmo.1 omm.mu N—w.m mem.1 Nun.mu _Nm.¢m mp Nm¢.mwwn ¢N¢.Ppu mo~.m w~m.w ccm.e1 www.mpu Fem.m weo.~ omN.Ppu mm¢.1 Nmm.u mmm.m mmm.mu me.op1 Num.o—u No¢.1 mnm.m up one. mmo. mmo. moo. app. mmo. Foo. eoo.- uno.1 mmo. mmo.1 mmo.1 “no.1 «no.1 ooo.u meo.1 mmo.1 mum.omm opo.pu Poo.u omm.¢ owo.wu oFN.NP NNN.F1 woe. www.mu mmo.1 om¢.1 neo.~1 mmo._1 mmm.1 www.mu mum.u NeF.N1 op www.mms mmm.mn mmm.P1 mw¢.¢u 5mm.m mmm. now.m m¢N.~1 omo.mu vpm.mu Fem.1 mem.pu m¢m.~ mom.mp ump.epn www.mu mn~.m1 .m mwo.~mmmm n mumeh mm:_m>:mmwm cem>oouz voucmwgwaxm ucmpmasoucH m>wpomepu< .m .o .H peeowp_eeeH “mane: awegaszz mpaeppmecz mpnawpmm “man peemappeeewea o>wmpaamm exam eeeem cmucmwgmaxmcH acouoaeou .mp .m_ .v— .m. .NP .FP .op 03 FNMQ‘LDKONG) mummucoo Aeezeapeoev m_ memee 92 mum.mpom mmm.woom mmm.mmme mmo.mmom mpm.mopm n¢v.m~mm Pmp.mmow noo.ommom mmzpm>cmmwm omo.m pmm.m mmn.ou pmm.omu www.cp ooo.e. ppm.mm1 mmm.mm vmocmwgmaxm .mp mmm.vp Pvm.wu an.~P1 mmp.m~1 one. o¢o.mp mm~.op mmm.oou “cmpmasoucH .vp mpm.u wmn.p Nm¢.opu ump.o_ mom.mm www.mm mmo.FN www.mF m>wpumgup< .m_ upo.m1 www.cu mm~.¢ moo.mu www.mu vmm.o me.m m¢o.mm .m .o .H .NF mm~.m Nem.wu omm.uu emm.mp eow.~mn me.mp mou.mpu emo.mm ucmmwppmucH .FF mom. m¢~.¢~1 mum.m1 mpm.om mmm.np mum.Pu mum.¢mu www.men nuanc: .op mwm.omu Num.n_1 mem.mn www.mpu www.muu Fm¢.¢mu voo.¢p owm.mpu amena531 .m mFm.m~n nop.w mmm.m~ mwm.o~1 mN¢.Ppu Nmo.¢m mom.m~1 mm~.¢m1 mpnmrpwgcn .m mmu.mp me.ppu sen.wpn owm.mp omm.m mmp.m~n mmm.mp mwe.n¢ mpnmPPmm .n Nmm.~1 mm~.m mNN.FP mum.Pmu mpm.ppu mm~.¢u nw~.¢m man.o¢ pmso .m moo.mu ¢m~.¢ emm.mp mov.m_1 mow.mm Nmm.m_u mm¢.w om—.qu pcmmwp—mucwc: .m cow.¢~ FNN.0N ¢o~.mp1 moo.m~1 ¢¢F.mpu oo¢.omu woo.mp1 mmp.¢¢1 m>wmpznmm .v NNN.w1 mmF.Fm mew.um mum.e mo¢.NP onF.mn www.mu who.mpu zzdm sugwm .m meo.ppu PmN.P~1 om~.¢ mmn.o~ www.mmu www.mu mmn.o¢ mwm.N¢1 umucmwgmaxmcm .N umn.opi mmm.m Fmo.n~ emu.m~ mN~.m_u mom.umu P~¢.om1 mpo.mm pcmumanu .— m .N. 1o. w. w m 1N. 1_ flag muh420 :ummmm mp m4mcmmwm cgm>oouz .o_ umucmwgwaxm .mp pneumasoucH .ep m>wpomcpu< .m_ .m .o .H .NF beemappeeeH .__ umanco .op xmgca531 .o mpamwpmgco .m mFQmwpmm .m pmzo .o beemw_Pepeee= .m m>wmpsamm .3 exam cuewm .m umucmmgmaxmca .N ucmpmqsoo .F Mag “emacwpcoov my m4mcmowm opm.mp Nmm.N~ oep.mm1 me~.on opp.mpu oo¢.m moo.p mmm.~Fu cgm>oouz .op ¢m¢.mp mom.o oom.¢pu mNo.m1 mo~.op opo.¢~ Nm¢.omu oom.mv omucmpemaxm .mp eom.n~ opo.o1 mem.u moo.¢Fu www.mu oom.- soo.mp ooo.¢on pcwumasoucm .ep omp.opu moo.m omm.¢u nmo.m moo.eu Nom.om oeo.o~ omo.¢~ m>wuumeuu< .m— mon.mu ooo.m num.~ opo.o omo.opu opo.m moo.m onp.mo .m .o .H .NF ooo.e ~N~.~—n Fem.mp nnp.ou omu.emn mF¢.N wen.m—u mop.mv ucmowppmpcm .FP omm.u omp.ou ooN.mpu oom.mm om~.~pu oom.o ovo.omu noo.m¢1 “mane: .o_ o~e.- woo.p1 opm.om oom.np1 oom.ou oeo.np Foo.¢pu mum.om1 magnaszz .o eoo.o1 omo.m omm.¢m omo.o1 oom.pp Nn¢.~p Nuo.N_1 opm.~mu mpamw—mgco .o mmo.PP moo.mu om¢.-1 mpo.op mom. Foo.pmu meu.¢p oo~.m¢ mpnwwpmm .n nmp.m mmo.~ omm.o omm.mmn oom.mm umo.-n nm¢.mm umn.mm umao .o po~.mF1 oe~.~p on.NN1 mon.o emo.¢m o~¢.¢n mum.o~ mm~.nmu pcwow—pmucpco .m sum.m Poo.m oem.su oom.¢~1 pmp._1 mo~.om1 www.mmu ooo.o¢1 m>wmpaamm .c non.uu «No.F Foo.nm omF.m1 mpo.ou npu.m Foo.m1 moo.omu chem cugwm .m muo.ou mom.m~1 o¢m.mp coo.o omn.omu opo.N~1 num.o¢ mu¢.m¢1 umucmpgmaxmcm .N onp.o1 owp. epF.op NNN.¢ mom.m oom.mmu _no.¢~1 ¢m~.om pcmpmneoo .P m M m m 1m m m 1P. g mmhHhHmom om moocmmwm ccm>oouz .o_ coucwwgqum .m— pcmumgeoucu .ep m>Ppumguu< .mp .m .o .H .N_ eeemPPFeHeH ._F “mane: .op xmgga521 .m mpnm_pmcc: .o mFQmwpmm .m pmao .o peema__epe_e= .m m>wmpaamm .q exam cuewm .m neocowgmaxmcH .N ucmpmasoo .F mammmmmm Aomzcwuzoov om mom~hcoovm cgm>oooz .op umucmpemaxm .mp pneumasouca .cp m>Ppumepp< .mp .m .o .H .NP beemap_epeH .FP “wage: .ep .8; :95: mpnmwpmcco mpnmppom gnaw peemr__epepe= m>wmpsqmm exam eeeam umucmwemaxmcH pcmumgsoo OS F-wa'l-O‘ONQ mummocoo 97 omo.omomu ~»o.mn mom.o»u o»o.¢ mew.» oo¢.mu eoo.m»u o»».o m~¢.mu m»o.o~ emo.- mm».o» m»m.»»1 moo.N »N».ou ~oo.o»n mm».~ o» o»o.»~oms nmo.mn emm.ou uo».-: o»~.»u »»».m omm.o» moo.o» m»¢.ou o»~.¢ om».o oNo.»» em».»» mm».¢u om».o1 omo.mu o»».o~- m» ¢»¢.»mo»u oge._ www.mm- eo~.mg om~.~- ngm.g mom.gg _om.~1 gm~.~ mne.~- emo.F mgN.~- ega.mg gag. Nmo.m ggg.om- Fgg.~g q» »o¢.omo1 mmo. oco.~u mm~.¢ o»~.» eem.ou omm.»: om~.¢» weo.mu mm».o»u »»o.~»n »oo.»» No~.mu »m».mn cwm.eu ooo.1 »oo.» o» omo.»»»1 mm».mn mom.» »o». oo».~- um».»1 moo.o omo.»n co».m1 emu.» om».m mom. mom. om».»s No».mu moo.u m»m.u N» omo.u »eo.n »»o. Nmo.u ooo.1 ooo.u moo.u »oo. e»o.1 oeo.u meo. moo. mo». »~o. moo. »¢o.1 eeo. vo».»m¢» nem.mu ooo.u oo¢.m» »mo.—»n Nem.MN m»o.m m»m. »»».m1 n»».» www.cn mup.» »Nm.m»1 mew.» om».ou emu. Nem.¢»n o» mum.»oo» ouo.»»u umo.»u m¢¢.-1 »oo.m oom.o moo.¢~1 mmo.»~ oo¢.o~1 »om.¢»1 m»o.¢» moo.¢»1 Noo.m~ moo.m »o».m»1 moo.¢»1 m»¢.m» .m mo».mm»»o u mung» mm:»n>:mmwm cgm>oouz .n» nmucmwgmaxm .m» pcmpmgoufi .3 m>wuunguu< .m» .m .u .H .~» nemmgggepeg .FF wanna: .o» xmgcaszz .m m»an»»mgco .m mpnnvpmm .» pmsn .n pcmmp»»mucw:: .m m>wm»=nmx .u gxnm cug»m .m nmucmwgmaxmcH .N ucmamasou .» g geesegueoeo _N mum> all CON TO POS Y—Z PLONE Figure 7. Three-dimensional representation of control group-positive induction group comparison. 1C 1‘. L/ 100 X—-‘( PLONE > '1; <:::: y’;3%;:,1{’is x—z PLRNE I” \11 Figure 8. Three-dimensional representation of control group-negative induction group comparison. 101 three dimensional representations illustrate Bayh's movement. From the three dimensional representation of the Speech only group comparison, it would appear that Bayh moved toward more negative trait descriptors than positive. It appears, in particular, that he moved toward incompetent (14). In the positive induction group comparison (Figure 7), it also appears that Bayh moved toward "incompetent" (l4) and repulsive (4) as well. In Figure 8, the three dimensional representation of the negative induction group comparison, it appears as though Bayh moved towards "Ideal Credible Source“ (12), or, at least, toward some of the positive trait descriptors. Caution, however, is suggested in attempting to infer too much from a three dimensional representation. We know that in the positive induc- tion group, Bayh moved closer to "Ideal Credible Source," and that he moved away from "Ideal Credible Source" in the negative induction group. These movements are not evident in the figures and a procedure is needed to assess Bayh's movement through all dimensions of the spaces. Before presenting this procedure, it should be pointed out that while the message used in this study was based on a pretest sample, I shall make our comparisons of the predicted motion and obtained motion using the control group to derive the predicted motion (the resultant vector). Some changes were obtained between this solution and the solution originally obtained for the pretest sample. The reason for these differences may be attributed to the fact that the means trimming computer program was not written before the posttest questionnaires were constructed. When the means were trimmed for the pre- test data and used in the procedure for obtaining messages, the resultant vector representing the four-pair message solution used in the study was 102 different than as previously presented (Chapter II). In the revised so- lution (after means trimming) the correlation between the target-ideal point vector (Rm) and the resultant vector (R1) was .932; an angle of 21.23 degrees. This varies from the solution presented earlier based on untrimmed means where the correlation was 1.00 (an angle of 0.99 de- grees). In the solution obtained with the trimmed control group means, the correlation between the resultant vector and the target-ideal point vector was .88 (an angle of 27.8 degrees). The difference between the solutions for the trimmed control group means and the trimmed pretest means may be attributed to small changes in either Humphrey's location or Bayh's location in the space. However, this change could not have been large or the correlation between the resultant and target-ideal point vectors would have been appreciatively reduced. It is not important, in the assessment of message effectiveness, whether or not this resultant vector is the optimal message. What is crucial is to show that Bayh's movement approximates this vector. To calculate a correlation between the obtained Bayh motion and pre- dicted motion, the control group and each of the three posttests groups were rotated together with the thirteen theoretically stable concepts Specified as stable (see Woelfel gt_pl,, 1975), and Bayh's time one loca- tion was subtracted out from the time two data. Functionally, these two steps locate the two configurations at a stable concepts centroid and then perf0rms a translation of the time two configuration (each of the posttest configurations) to a centroid which is comparable to locating the coordinates of Bayh at time one at the centroid (a procedure utilized when deriving the message). Vectors are then compared between Bayh's time two location (his motion vector) and the (time one) resultant vector. 103 (For elaboration, see Woelfel gp_pl,, 1976). Before presenting these correlations, I shall review the set of findings II expect. According to the underlying assumption upon which procedures used in this study are based, §s should combine the fOur con- cepts Bayh was associated with and Bayh (in the positive induction group) should move approximately along the resultant vector. In the speech only group, I do not necessarily expect Bayh's motion to parallel this resultant vector. (Bayh will move, however, due to some inferences §§ make from the speech.) Further, Bayh should move in a direction approximately opposite to the resultant vector in the negative induction group. (The reason a resultant vector with the concepts "dissimilar to Humphrey," "unjust," "inexperienced," and "incompetent" was not used for the correlation of Bayh's motion in the negative induction group is be- cause the message disassociated Bayh and Humphrey. The computer program written to assess message effectiveness cannot be utilized in the case of a disassociated linkage.) Therefore, the expected correlations are positive, zero and negative in the positive induction, speech only and negative induction groups, respectively. Table 22 presents the correlation between Bayh's motion vector and the resultant vector for each of the three experimental posttest groups rotated to the control group configuration specifying the thirteen stable concepts as stable. In addition, this table presents the correlations between Bayh's motion vector and each of the single concept vectors -- to indicate each attribution's relative effect. In the control group-Speech only group comparison, Bayh's motion vector and the four pair resultant vector correlated .273. The highest correlation for a single-concept vector (“competent") was .501 (an angle of 59.90 degrees). Thus, it 104 TABLE 22 CORRELATIONS AND ANGLES FOR BAYH'S MOTION AND THE RESULTANT VECTOR* Control- Control- Control- Message Speech Positive Negative competent, experienced, .297 .495 .687 Humphrey and just 72.72 60.33 46.62 competent .501 .579 .862 59.90 54.58 30.47 just .355 .621 .773 69.21 51.60 39.41 Humphrey .228 .615 .648 76.82 52.05 49.60 experienced .295 .557 .634 72.85 56.16 50.62 distance moved: 86.18 63.41 89.88 * The first number is the correlation; the second number is the angle. 105 appears that Ss inferred some degree of competency about Birch Bayh based on the speech. Other correlations, in order of magnitude, were .355 ("just," an angle of 69.21), .295 ("experienced," an angle of 72.85 de- grees) and .228 ("Humphrey," an angle of 76.82). Finally, Bayh moved 63.41 units due to the speech. In the positive induction group comparison, Bayh's motion vector correlated .495 with the four-pair resultant vector (an angle of 60.33). In order of magnitude, Bayh moved along the “just" vector (r=.621, an angle of 51.60 degrees), the "Humphrey“ vector (r=.615, an angle of 52.05 degrees), the "competent" vector (r=.579, an angle of 54.58 degrees) and the "experienced" vector (r=.557, an angle of 61.90 degrees). In the positive induction group, Bayh moved 86.18 units. In the negative induction group, Bayh's motion vector correlated .687 (46.62 degrees) with the four-pair resultant vector. In order of magnitude, Bayh moved along the "competent" vector (r=.862, an angle of 30.47), the "just“ vector (r=.773, an angle of 39.14 degrees), the "Humphrey" vector (r=.648, an angle of 49.60), and the "experienced" vec- tor (r=.634, an angle of 50.62 degrees). Bayh moved 89.88 units in the negative induction group configuration. The correlation between Bayh's motion vector and the predicted mo- tion vector in the negative induction group is remarkably contrary to the predicted correlation. A moderate to high negative correlation was predicted for the Bayh motion-resultant vector correlation, but the ob- tained correlation was .687. Further, there was some degree to which the Bayh motion was skewed, or biased, towards "competent“ and “just“--these correlations (.862 and .773, respectively) were much higher than the others. The failure to support this implicit hypothesis may be due to 106 two causes: effects of the persuasive speech (which we have found to have moderate effects on the Ss' perceptions of Birch Bayh) and/or to the fact that associating, or disassociating, Bayh with Humphrey produced effects that had not been previously considered. This second point needs further elaboration. Let us reassess the starting assumption: when two concepts in the space are associated (formally, when they are linked in an assertion of the form "x" is "y") they converge relative to one another along the vector connecting them. Thus if a public figure is associated with a stable concept, we expect the public figure to move toward the stable concept. Similarly, if asso- ciated with four stable concepts, the public figure Should move in a direction which approximates their resultant vector (or perhaps skewed towards one of the stable concepts), and the stable concepts should re- main stable. However, little is known about the effects of associating (or disassociating) Bayh with Humphrey--particularly if Humphrey moved in the space (which seems plausible given the observed motions illustrated in Figures 6, 7 and 8). Thus, in order to shed light on the Bayh motion vector-resultant vector correlations, we shall first assess how Humphrey moved in the space and how this motion affected Bayh's motion. In view of the three dimensional representations (Figures 6, 7 and 8), it appears that Humphrey (and McGovern) moved toward the negative por- tion of the semantic space. To assess Humphrey's motion, however, across all sixteen dimensions, his motion vector was correlated with each of the six negative trait descriptors. These correlations are presented in Table 23 for each of the three control-posttest group comparisons. 107 TABLE 23 CORRELATIONS AND ANGLES FOR HUMPHREY'S MOTION AND THE SIX NEGATIVE TRAIT DESCRIPTORS* Control- Control— Control- Message Speech Positive Negative inexperienced .767 .604 .409 39.92 52.86 65.83 repulsive .508 .803 .529 53.24 36.54 58.03 unintelligent .554 .604 .339 57.01 45.85 70.19 unreliable .450 .935 .355 63.22 20.70 69.19 unjust .322 .739 .347 71.21 42.32 69.80 incompetent .412 .856 .439 65.65 31.28 63.95 distance moved: 78.36 53.24 39.20 * The first number is the correlation; the second number is the angle. 1C /\ L, 108 Evidently, the Bayh attributed speech had drastic effects on Hum- phrey's location in the space. In the control group-speech only group comparison (the first column in Table 23), Humphrey's motion vector cor- related, in order of magnitude, with: "inexperienced" (r=.767, an angle of 39.92 degrees); "unintelligent" (r=.554, an angle of 57.01 degrees); "repulsive" (r=.508, an angle of 53.24 degrees); "unreliable" (r=.45, an angle of 63.22 degrees); "incompetent“ (r=.412, an angle of 65.65 de- grees); and "unjust" (r=.322, an angle of 71.21 degrees). These correl- ations indicate that while Bayh moved along the "competent" and "just" vectors due to the Bayh attributed Speech, Humphrey moved along the "inexperienced," "unintelligent" and "repulsive" vectors. When Bayh was described as "Similar to Humphrey," "incompetent," "experienced" and "just," Bayh moved along the "just," "Humphrey" and "competent" vectors (see second column of Table 22). However, Humphrey moved along the "unreliable" vector (r=.935, an angle of 20.70 degrees). Indeed, all the Humphrey motion vector-Single concept vector correlations are fairly high in the control group-positive induction group comparison (see second column of Table 23). When Bayh was described as "dissimilar to Humphrey," "inexperienced," "incompetent" and "unjust," Bayh moved along the "competent" and "just" vectors (see third column of Table 22). However, Humphrey moved along the "repulsive" and "incompetent" vectors (correlations of .529 and .439, angles of 58.03 and 63.95 degrees, respectively). Thus it would appear that Humphrey moved due to the Bayh attributed speech and that the associative and disassociative linkages made between Bayh and Humphrey also had considerable effects on Humphrey's location in the space. The important conclusions to draw from the correlations L. 109 presented in Table 23 are that (1) Humphrey moved toward the negative part of the semantic space, due to the Bayh attributed speech; (2) his movement is consistently in the direction of the negative part of the semantic space (across negative traits); and (3) he moved toward the negative part of the semantic space both when he was associatively linked with Bayh and when he was disassociatively linked with Bayh (across conditions). We shall now want to assess how Bayh's motion related to Humphrey's motion. To do this, correlations were computed between Bayh's motion vector and Humphrey's motion vector for the three comparisons. These motion vector correlations are presented in Table 24. First, it should be noted that a moderate to high positive correlation indicates that the concepts' motions through the space are parallel and in the same direction. A moderate to high negative correlation indicates that the concepts' motions through the space are parallel and opposite in direc- tion. In the control group-speech only group, Bayh's motion and Humphrey's motion correlate .204 (an angle of 78.23 degrees), not a correlation sub- stantial enough to say that there is a concrete relationship. However, in the case where Bayh was described as "similar to Humphrey," "just," “competent," and “experienced," Bayh's motion vector correlated .419 (an angle of 65.19 degrees) with Humphrey's motion. Thus, it appears that the associative linkage between the two public figures succeeded, in part, in a nearly parallel movement in the same direction. Two statements can be made in drawing implications of this result of interrelated movement. First, the original purpose of the study was to demonstrate that a resultant vector can be constructed from time one e, H \ \3 1 f 1 I‘VE“. .- '. - - 110 data, a message devised utilizing this resultant vector, and a concept can be moved along this resultant vector. The Bayh-resultant vector correlations presented in Table 22 correlated Bayh's time two location (motion vector) with the resultant at time one. While it has been demonstrated that the theoretically stable concepts are indeed stable, it has also been demonstrated that Humphrey moved appreciatively. Second, since Humphrey's motion in the control group-positive in- duction group is toward the negative part of the space (see second col- umn of Table 23) and Bayh's motion vector correlated .419 with Hum- phrey's motion vector, it seems reasonable to conclude that the asso- ciative linkage with Humphrey attenuated Bayh's motion along the pre- dicted resultant vector. When Bayh was described as "dissimilar to Humphrey," “incompetent,“ "inexperienced" and "unjust," Bayh's motion vector correlated -.449 (an angle of 116.74 degrees) with Humphrey's motion vector. This indicates that the disassociative linkage between Bayh and Humphrey resulted in motion vectors that were opposite in direction and formed an obtuse angle. Thus, while Humphrey moved toward the negative part of the semantic space (see third column of Table 23), Bayh's motion, in part, is based on movement away from Humphrey's motion. Thus, Bayh moved toward the positive part of the semantic space as if the other attributions (“in- competent," "inexperienced" and “unjust") had no effect. While a cor- relation of -.449 is not substantially high, it suggests sufficient effects to have moved Bayh considerably different from the hypothe- sized motion. Note that the correlations presented thus far are based on using the control group configuration as the baseline f0r comparison with the 111 TABLE 24 CORRELATIONS BETWEEN HUMPHREY AND BAYH'S MOTION VECTORS* Control- Control- Control- Speech Positive Negative .204 .419 -.449 78.23 65.19 116.73 * The first number is the correlation; the second number is the angle. ‘ \ J 1. 112 three posttest group configurations. In both the positive induction group and the negative induction group, the speech was delivered after the induction message. The meaning of the speech, therefore, is dif- ferent from the meaning of the speech in the speech only group because the meanings (locations in the spaces) of several of the concepts in the multidimensional space had been altered by the induction message prior to the delivery of the speech. Therefore, by using the control group configuration as the baseline fer comparisons, we have observed the effects of the Speech, the induction message and the interaction between speech and induction message. The effects of the speech and the induction messages are confounded. There may be linear effects of the speech, linear effects of the induction message and non-linear (interaction effects) between Speech and induction message, and there is no positive induction only group or a negative induction only group in which to provide full explanations of the obtained motions. However, if the speech only group configuration were to be used as the baseline fer comparing the positive induction group and negative induction group configurations, it would be possible to assess the lin- ear effects of the induction messages plus the Speech-induction message interaction effects. In a sense, this would be eliminating the linear effects of the speech. I. shall briefly make this comparison, but first note that Since Humphrey moved toward the negative part of the space due to the speech and is a concept used in the f0ur-pair result- ant vector, it would seem plausible that this motion would have substan- tial effects on the resultant vector obtained in the speech only group configuration. To assess this, the procedure used to derive the result- ant vector from the control group configuration (see Chapter II) was 113 used on the speech only group configuration. The four-pair resultant vector correlated .927 (an angle of 21.96 degrees) with the target- ideal point vector. Thus the four-pair resultant vector is comparable when based on the Speech only group configuration. To assess the motions obtained after eliminating the linear effects of the speech, the positive induction group and the negative induction group configurations were (separately) rotated to the Speech only group configuration, and the Bayh motion vector-resultant vector correlations were computed. These correlations are presented in Table 25. Due to the linear effects of the positive induction message and the speech- positive induction message interaction, Bayh's motion vector correlated .660 with the four-pair resultant vector. The correlations between Bayh's motion vector and the Single concept vectors are, in order of magnitude: “competent" (r=.672, an angle of 47.77 degrees), “just“ (r=.643, an angle of 49.98 degrees), "experienced" (r=.622, an angle of 51.53 degrees) and "Humphrey" (r=.565, an angle of 55.61 degrees). In the speech only-negative induction group comparison, Bayh's motion vector correlated .325 with the resultant vector (an angle of 71.04 de- grees). The correlations between Bayh's motion vector and the Single concept vectors are, in order of magnitude: "Humphrey” (r=.549, an angle of 56.72 degrees), "competent" (r=.445, an angle of 63.57 degrees), "experienced" (r=.360, an angle of 68.89 degrees) and "just" (r=.228, an angle of 76.83 degrees). The correlation between the Bayh motion vector and the resultant vector in the speech only-positive induction group comparison indicates that the positive induction message did indeed move Bayh along the re- sultant vector. In addition, the Bayh motion vector-Humphrey motion \ ’5‘ 114 TABLE 25 CORRELATIONS AND ANGLES BETWEEN THE BAYH MOTION VECTOR AND THE RESULTANT VECTOR IN THE SPEECH ONLY-POSITIVE INDUCTION MESSAGE COMPARISON AND THE SPEECH ONLY-NEGATIVE INDUCTION MESSAGE COMPARISON * Speech- Speech- Message Positive Negative competent, experienced, .660 .325 Humphrey and just 48.71 71.04 competent .672 .445 47.77 63.57 just .643 .228 49.98 76.83 Humphrey .565 .548 55.61 56.72 experienced .622 .360 51.53 68.89 distance moved: 63.38 64.07 * The first number is the correlation; the second number is the angle. 115 vector correlation decreased to .061 (Table 26). Thus, when using the control group configuration as the baseline, the Bayh motion vector- resultant vector correlated only .495, but when the linear effects of the speech are eliminated, the correlation increases to .660. There- fore, the linear effects of the speech apparently operate to: (1) re- duce the amount of parallel movement between Bayh and Humphrey (their movements are orthogonal--suggesting that the associative linkage made in the positive induction group had little effect and/or that this effect presumably is counterbalanced by the effect of the Speech- positive induction message interaction); and (2) attenuate the Bayh motion vector-resultant vector correlation. Unfortunately, the speech- positive induction interaction effects cannot be eliminated with the present set of data. Bayh also moved along the resultant vector in the speech only- negative induction group comparison. This correlation is attenuated after eliminating the linear effects of the speech (see third column in Table 22). Nonetheless, the correlations presented in Table 25 do indicate movement contrary to the hypothesized movement. This may be explained by the fact that there exists a strong relationship between the Bayh motion vector and the Humphrey motion vector (r=-.704, an angle of 134.78 degrees; see Table 26). Evidently, the disassociative linkage made between Bayh and Humphrey in the negative induction mes- sage (plus interaction with the speech) resulted in Bayh and Humphrey moving in opposite and fairly parallel directions. 116 TABLE 26 CORRELATIONS AND ANGLES BETWEEN HUMPHREY AND BAYH'S MOTION VECTORS (SPEECH ONLY-POSITIVE INDUCTION GROUP COMPARISON AND SPEECH ONLY-NEGATIVE INDUCTION GROUP COMPARISON) * Speech- Speech- Positive Negative .061 -.704 86.50 134.78 * The first number is the correlation; the second is the angle. CHAPTER IV DISCUSSION The purposes of the study were: (1) to propose a multidimensional measurement model; (2) to hypothesize movement of a manipulated concept in the multidimensional space; and (3) to hypothesize that such move- ment would not only be predictable, but would induce higher and lower levels of credibility of the manipulated concept. The results of the study will be discussed in two sections: (1) Overview of the Measure- ment Model, and (2) Overview of the Effects of the Experimental Manipu- lation. Finally, we shall want to discuss directions for future research. Overview of the Measurement Model A multidimensional measurement model was proposed fer general use in person perception research. In proposing the model we have provided the fellowing set of definitions: (a) an attribute is a line segment be- tween points representing linguistic units which §§ perceive as semantic opposites; (b) a dimension refers to a reference line, orthogonal to all other dimensions, through the configuration of attribute end-points; and, (c) a cognitive ggppip_is a set of objects or concepts that naturalistic- ally possess some classificatory characteristic in common. Elements that reside in a domain can be objects or concepts (i.e., Presidents, friends, etc.) and linguistic units associated with such objects or used to describe such objects. 117 118 Application of the proposed measurement model is based on four assumptions: (1) Within a given domain, it is assumed that there exists a struc- ture; i.e., a formal set of relations among the linguistic units used to describe objects residing in the domain; (2) It is assumed that the meaning of a linguistic unit is deter- mined by its dissimilarity relations (physical separation in the Spatial representation) to all other concepts in the domain; (3) Within a given domain, it is assumed that a subset of linguistic units (descriptors) will bear gpgplg, linguistically determined, rela- tions to each other, providing a structure which is generally applied to other linguistic units (objects) within the domain; (4) It is assumed that §§ can be instructed to report ratio judgments of dissimilarities between linguistic units (descriptors and objects). The measurement model was developed in consideration of studies which have tested assumptions of the semantic differential (Anderson, 1970; Danes and Woelfel, 1975; Green and Goldfried, 1965; Gulliksen, 1957; Messick, 1957; Wishner, 1960) that revealed a need for a model possessing a more parsimonious and tenable set of assumptions. Considerable support was cited for the above four assumptions in Chapter I. Essentially, Assumption 1 states that some linguistic units will be relevant to a domain, and some will not. It is an empirical question quite often not answered (Hastorf pp 21,, 1958). In the present study, 96 trait adjectives were sampled from factor analytic measures of source credibility. Twelve adjectives, six sets of bipolars, were selected based on the consistency by which they occurred in the same category in a sorting task. Note that a potential source of error may stem from the s.» 119 use of "a close friend" as the reference person in the sorting task. In utilizing this reference person, we perhaps have crossed our own domain Specificity stipulation. However, at this point in time, it is not known to what degree the traits presently selected would not have been select- ed, given a change in reference person. Alternatively, future research may want to avoid using a sorting task in the fashion in which it was employed in the present study. This is for the obvious reason that the experimenter has too much of an in- fluence on the final results: s/he selected the initial set of linguis- tic units to be sorted. A very strong candidate for an alternative pro- cedure would be an application of the "attribute list“ notion (Phillips, in progress). The procedure would have 55 provide the attributes for the domain. The experimenter can either have Ss sort these attributes, use the more frequently mentioned attributes, or conduct a pretest of samples of attributes to select a maximum number of dimensionS--perhaps a maximum number of interpretable dimensions, if interpretability seems important. Assumption 2 merely states a basic underlying assumption of multidimensional scaling research (for examples, see Miller, 1967; 1969). The smaller the angle between two concepts, the more "similar" they are-~the more they overlap in meaning. Assumption 3 is fairly com- plex. It suggests the importance for defining the domain and raises the question concerning stability. Both issues are empirical questions-- what constitutes a domain; what constitutes stability? As argued in Chapter I, the domain specificity is an important prerequisite for sta- bility. Little is known about instability that results in moving from one domain to another (Hanno and Jones, 1973; Doherty, 1973), and the initial 120 work must be addressed to the question of what constitutes a domain. This may well take the form of hierarchical cluster analysis of occupa- tions. Then one can ask the question of stability: Are linguistic de- scriptors stable across a domain of ”national politicians" to a domain of "newscasters/entertainers," etc.? Nonetheless, in the present study twelve linguistic descriptors pur- ported to be terms generally applied in characterizing objects in the domain of "public figures," and presumably relevant to the credibility construct, were arrayed in a multidimensional configuration. Separate tests of stability were made and it was concluded that the array of de- scriptors was very stable. Specifically, (l) The multidimensional array of descriptors in each four posttest group correlated highly with the aggregate level configuration they presumably tap; (2) Each concept's location correlated highly between groups (with the possible, yet few, exceptions of "intelligent" and “just"); and, (3) The factor structure of the descriptors correlated highly in group comparisons. In addition, the fact that such concept stability and factor structure stability was obtained provides indirect, yet strong, evidence that Ss were instructed properly in providing ratio scaled distance estimates (Assumption 4). Attribute salience is a measure, internally provided by the model, which indicates how salient and important an attribute is in perceiving an object. Similar to Zajonc's (1969) construct of prominence, attribute salience is the ability of an attribute to represent or characterize the referent. This measure was found to correlate fairly high with a unidi- mensional measure of attribute importance. It was argued (Chapter I) that the attribute salience scores are analogous to weights. One may ex- pect that using a message centered around an attribute that is very 121 salient in characterizing the referent "Ideal Credible Source" would pro- duce effects different than using a message centered around an attribute end-point of a less salient attribute. In the present study, for example, one may wish to see if the more salient attribute end-point descriptors would have more effect in skewing (biasing) the motion of the public fig- ure (if primacy and/or recency effects are partialled out) than a less salient attribute end-point. This analysis was not done with the present sets of data for two reasons. First, the three attributes used in the messages ("just-unjust," "competent-incompetent," and "experienced-inex- perienced") were all fairly salient attributes in characterizing the "Ideal Credible Source" (Table 16). Second, and more important, the man- ipulated public figure's motion was too strongly affected by Humphrey's motion. Nonetheless, future research can assess the potential differen- tial effects of various salient attributes. Overview of the Effects of the Experimental Manipulations Procedures employed in the present study were designed to move the manipulated public figure towards the "Ideal Credible Source.” These procedures were successful in the positive induction group. They were not successful in inducing the hypothesized motion in the negative induc- tion group. The reason for failing to predict the manipulated pub1ic figure's motion is that a different, yet equally fundamental, psycholog- ical process operated jointly with the process originally assumed to operate. This process, the principle of mediated generalization (Tannen- baum, 1968), will be discussed below. Yet, because of the complexities inherent in the present set of data, I shall have to extend the predic- tions made by this principle. 122 The procedures employed in the present study were based on the assumption that if a public figure is associated with a set of stable concepts, the public figure Should move in a direction that approximates their resultant vector (or perhaps skewed towards one of the stable concepts due to primacy, recency or salience of an attribute). While, for the pretest group configuration and the control group configuration, it appeared that attributing similarities between Bayh and Humphrey would result in Bayh's moving along the resultant vector, results indi- cated that this associative linkage attenuated Bayh's movement along the resultant vector. While Bayh, to some degree, did move towards Humphrey's time one location (control group configuration), as hypothesized, his. movement was also interrelated with Humphrey's as Humphrey moved from time one (control) to time two (positive induction group configuration). When the linear effects of the speech were eliminated, the two candi- dates' motions were not interrelated, and Bayh's motion correlated higher with the resultant vector (based on the speech only group configuration). Further, from the control group and pretest group configurations, it appeared that disassociating Bayh and Humphrey (describing them as dis- similar) would tend to move Bayh away from the "Ideal Credible Source," away from the resultant vector. This was far from true. While Humphrey moved toward the negative portion of the semantic space, Bayh moved away from Humphrey (particularly when the linear effects of the speech were eliminated). Bayh moved in a direction opposite to that hypothesized. Thus, it is clear that in the present study that the observed ef- fects cannot be totally attributed to associating a public figure with four stable concepts. Effects have been found due to some other dynamic process, and I Shall consider what additional process, other than direct I . Q- I 123 attributions, that can account for the observed effects. Recent research of an application of Congruity Theory provides a starting point. Tannenbaum (1968) has argued that a single source-concept linkage consti- tutes a direct relationship bewteen two concepts and that a change in attitude toward one concept influences the relationship between both con- cepts, yielding an attitude change toward the second. He further argued for an extension of the Congruity Principle to applications of single source-multiple target communication situations. This extension can be clarified by presenting the fellowing study (discussed in Tannenbaum, 1968). Tannenbaum selected two previously neut- xpl, unassociated concepts, "Teaching Machines" and "Learning Theory," as well as a single neutral source. The source made assertions toward the concepts that both were positive (pp), positive towards "Teaching Machines" and negative towards “Learning Theory" (pn), negative towards "Teaching Machines" and positive towards "Learning Theory" (np) or negative towards both (nn). After a one-half-hour interval of irrelevant activity, one-half of the Ss received a message favorable towards "Teaching Machines" (P) and one-half of the Ss received a message which attacked "Teaching Machines" (N). Congruity Theory predicted favorable changes in attitude toward "Learning Theory" in the Ppp, an, an and Pnn conditions; unfavorable changes toward "Learning Theory" in the Ppn, Nnn, Npp and Pnp conditions. All predictions were supported. This principle, mediated generalization, asserts that if two concepts are linked (either associatively or disassociatively), and a subsequent induced change in one concept is achieved, there will be a change in the linked concept, even though this linked concept was not manipulated by the induced change of the first concept. It is possible to adapt this 124 principle to generate additional hypotheses about the manipulations used in the present study. In attempting to generalize the principle of mediated generalization to the present study, let us first assume that both Bayh and Humphrey are two "previously neutral, unassociated concepts." We shall later retract this obviously erroneous assumption. First, in the positive induction group, an associative link was created between Bayh and Humphrey and Bayh was described as "competent," "just" and "experienced." Thus, we would expect both Bayh and Humphrey to move toward these concepts. In the nega- tive induction group, a disassociative link was created between Humphrey and Bayh, and Bayh was described as "incompetent," "inexperienced" and "unjust.“ Humphrey should move away from the resultant vector fer these concepts. Thus, according to the principle of mediated generalization, adapted to the present study, Bayh should move away from resultant vector and Humphrey should move (approximately) along the resultant vector. The analysis of Humphrey's motion and Bayh's motion clearly indicates that the predictions derived above are not supported. In both induction groups, Bayh moved along the resultant vector, and Humphrey moved toward the negative linguistic descriptors. The obvious reason for this, and suggests a limitation on the set of situations applicable under the assump- tions of the principle of mediated generalization, is that Humphrey is not an initially neutral concept. While future research may well be dir- ected towards applications of the principle of mediated generalization with linkages between initially neutral source, the present results suggest that an old attitudinal object is different than a new attitudinal object. Since Bayh is a moderately familiar concept (he approximates a “neutral" concept), his motion Should be predictable, either by direct attribution 202k». , lb... 11.1 V 125 or via mediated generalization. But Humphrey is highly familiar to the §s, and his motion is quite contrary to what was expected. Humphrey's motion that was due to the Speech may be explained in two ways. First, note that the Bayh attributed speech impressed the §s and Bayh moved along the "competent" and "just" vectors. Humphrey may have moved towards the negative linguistic descriptors because he was con- trasted with the new impression of Bayh. This explanation, however, seems implausible because if the associative linkage between the two had any effects, Bayh gpg_Humphrey should have moved toward positive linguistic descriptors. Instead, Humphrey moved very closely along the "incompetent" vector. Indeed, when eliminating the linear effects of the speech, Bayh more closely paralleled the resultant vector, suggesting that the associa- tive link did more to hurt Bayh than help Humphrey. Alternatively, it may be the case that the Bayh attributed speech contained in it contents which activated old information the Ss possessed about Humphrey. If this explanation is true, and future research should assess its validity, then the variations in Humphrey's movement in the Space toward negative descriptors (across the three posttest experimental groupS--Table 16) may be due either to the linkages made with Bayh or may be due to different types ()f information about Humphrey being activated in different groups of §s in the posttest groups. However, since there is fairly good evidence that §s were adequately randomized into groups (both demographic information and pretest on attitudes on the topic), this sec- ond explanation of Humphrey's movement seems rather implausible. It would appear that Humphrey moved toward the negative descriptors due to the ac- tivation of old information, and that the various Bayh-linkages resulted in variations in motions which were consistently in the direction of the 126 negative descriptors. Thus, it would appear that Humphrey's motion in the space and the various linkages made between Bayh and Humphrey resulted in quite differ- ent effects on Bayh's motion. When the linear effects of the speech were eliminated, there was good evidence that the interrelatedness of the Bayh-Humphrey motion was attenuated and the degree to which Bayh moved along the resultant vector increased. In the negative induction group, once the linear effects of the speech were eliminated, Bayh and Humphrey's motions were highly interrelated and Bayh's motions along the resultant vector was attenuated (in comparison with the correlation obtained when including the linear effects of the speech). These results suggest that the type of linkage, associative or disassociative, operated considerably differently. Apparently, the disassociative linkage was considerably stronger than the associative linkage, so much so that the other attribu- tions ("incompetent," "unjust" and "inexperienced“) had no effect whatso- ever. In relation to source credibility, procedures employed in the pres- ent study did relocate Bayh at a point closer to "Ideal Credible Source" in the positive induction group, and further away from this point in the negative induction group (see "Hypothesis" section). This, however, takes into account only the mean interpoint distances between Bayh and the "Ideal Credible Source." When the direction of movement is taken into account, Bayh did not move away from the "Ideal Credible Source"; nor the resultant vector. Indeed, Bayh moved toward "competent" and "just” in the negative induction group configuration -- towards a new location in the positive part of the space. However, the interpoint distance between Bayh and Ideal Credible Source is larger in the negative induction group 127 configuration than in the positive induction group and, if McLaughlin's (1975) argument is correct, Bayh should be more credible in the positive induction group. Unfortunately, this hypothesis was not adequately tested in the present study because the persuasive speech was persuasive by itself and the credibility inductions had little additional impact. As of yet, McLaughlin's recommendation is still untested. Future research may want to test the hypothesis that "the public figure perceived as closer to the 'Ideal Credible Source' will stimulate more attitude change than messages attributed to a public figure who is further away from 'Ideal Credible Source'" in a study which simply locates public figures in the space, utilizes no credibility manipulations and utilizes a speech which is not as highly persuasive as the one employed in the present study. Directions for Future Research Concerning the proposed model, three questions need to be investigated. First, I have discussed the problem of domain specificity and, interrelated with it, the question of stability. Utilizing an adaptation of the "attri- bute list" (Phillips, in progress), attributes relevant to various domains can be generated and arrayed in a multidimensional configuration. Varia- tions on that result from moving from one domain to another can be observed by assessing the overlap (or lack of overlap) in attributes obtained in the list, or by how an attribute (or set of attributes) not relevant to a particular domain load(s) differently in various domain representations or exhibit instability. Further, one can obtain an independent assessment of attribute relevance for a domain and test Doherty's (1973) implicit hy- pothesis that attribute end-points of attributes relevant to a domain ex- pand when moving from one domain to another. 128 Second, the stability of factor structure needs further investiga- tion. Earlier research (Barnett, 1972) reported that metric multidimen- sional configurations become fairly stable when the number of respondents reaches 70. However, in the present study, we have sampled concepts within a particular domain and have obtained fairly high Split-half fac- tor structure stability with sample sizes far smaller than 70. Future research is needed to test the stability of factors across domains varying sample sizes. Such a study can provide information concerning the re- quired number of subjects needed for conducting research in specific do- mains. Such infbrmation would have been extremely valuable in designing the present study. If we had previous knowledge concerning sample sizes, it may have been possible to use fewer subjects with each condition and add positive induction only and negative induction only groups. Finally, as discussed earlier in the chapter, research should be conducted on the differential effects of utilizing salient versus non- salient attribute end-points. While the model has been demonstrated to be a useful means for assess- ing structure, generating messages and assessing the effects of messages, the results of the procedures employed in moving Birch Bayh towards "Ideal Credible Source" generate far more questions than answers. As mentioned above, McLaughlin's (1975) hypothesis is still untested. In terms of man- ipulating credibility, research is needed to provide a test of the basic assumptions of the procedures employed--studies are needed to test the sole effects of direct attributions of stable concepts (without speech effects, or construction of linkages with other person-objects). Given that Such procedures are successful, research can be directed at three additional areas: (1) varying messages that are based on one-pair, two- 129 pair, three-pair and four-pair resultant vectors; (2) varying (language) intensity in the messages; and, most importantly, (3) utilizing the proce- dures outlined here (and further detailed in Woelfel g£_gl,, 1976) in an on-going campaign with messages generated and effects tested over time. Further, results of the present study indicate that there are dram- atic effects obtained when linkages are made between person-objects. While it may be highly profitable (from the standpoint of Congruity Theory) to assess how previously neutral concepts can be linked and mani- pulated in ways Specifically predicted by the principle of mediated gen- eralization, such results appear to lack substantial generalizability. It is far more often the case that public figures are described in terms of other public figures, run on the names and successes of past public figures and are disassociated and associated with other, current, public figures. Thus, additional research that creates linkages between unfa- miliar and familiar public figures, or two (or more) familiar public fig- ures would be pragmatically appealing. Such research, at first, would be exploratory, particularly on how old information about an old, familiar, attitudinal object is activated by a speech delivered by another public figure (resulting in different impressions of both figures), but would add tremendously to the ability of predicting and manipulating elements in the social environment. A <\\ ‘ l d tel...“ __ .34 .. .. BIBLIOGRAPHY BIBLIOGRAPHY Anderson, A. 8. Structure of semantic space. In E. F. Borgatta and G. E. Bohrnstedt (eds.), Sociolo ical Methodolo 1970, San Francisco: Jossey-Bass, Inc., 1970. PP. 308-325. Anderson, A. 8. Brief Report: the effect of aggregation on non-metric multidimensional scaling solutions. Multivariate Behavioral Re- search, July 1970, pp. 369-373. Asch, S. E. Forming impressions of personality. Journal of Abnormal and Social Psycholggy, 1946, 41; 258-290. Barnett, G. A. "Reliability and metric multidimensional scaling." Un- published research report, East Lansing, Michigan, Michigan State University, 1972. Barnett, G. A. A Paradigm for the Measurement of Meaning. Unpublished preliminary paper, Department of Communication, Michigan State University, 1975. Barnett, G. A., Serota, K. 8., and Taylor, J. A. A Method fer Political Communication Research. Paper presented at the Annual Convention of the Association for Education in Journalism, San Diego, California, August 1974. Berlo, 0., Lemert, J., and Mertz, R. Dimensions for Evaluating the Acceptability of Message Sources. Public Opinion Quarterly, 1969, 33, 563-576. Bowers, J. W., and Phillips, W. A. A note on the generality of source- credibility scales. Speech Monographs, 1967, 34; 185-186. Bruner, J. S. and Taguiri, R. The perception of people. In G. Lindsey (ed.), Handbook of Social Psychology, Vol. 2, Cambridge: Addison- Wesley, 1954, pp. 634-654. Bruner, J. S., Shapiro, 0., and Taguiri, R. The meaning of traits in isolation and in combination. In R. Taguiri and L. Petrullo (eds.), Person Perception and Interpersonal Behavior. Stanfbrd: Stanford University Press, 1958, pp. 121-147. "Carroll, J. 0. Individual differences and multidimensional scaling. In R. N. Shepard, A. K. Romney, and S. Nerlove (eds.), Multidimensional Scalin : Theor and Application in the Behavioral Sciences. (Volume II AppiicationsI. New York: Seminar Press, 1972. Carroll, J. 0., and Change, J. J. A general index of nonlinear correla- tion and its application to the interpretation of multidimensional scaling solutions. American Psycholpgist, 1964, 19: 540. 130 131 Cattell, R. 8. Factor Analysis. New York: Harper and Brothers, 1952. Cliff, N. Orthogonal rotation to congruence. Psychometrika, 1966, 31: 32-42. Cody, M., Marlier, J., and Woelfel, J. A Reconceptualization of Multiple Attribute Measurement: The Location of Unidimensional Scales in an N-Dimensional Space. Presented at the Eighth Annual Math Psychology Meetings, Purdue University, W. Lafayette, Indiana, August 1975. Coombs, C. H. Psychological scaling without a unit of measurement. Psychological Review, 1950, 51, 148-158. Coombs, C. H. A Theory of Data. New York: Wiley, 1964. Coombs, C. H., Dawes, R. M., and Tverskey, A. Mathematical Psychology, Englewood Cliffs, New Jersey: Prentice-Hall, 1970. Craig, R. T. Models of Cognition, Models of Messages and Theories of Communication Effects: Spacial and Network Paradigms. Preliminary paper, Department of Communication, Michigan State University, 1974. Cronbach, L. J. Processes affecting scores on 'understanding of others' and 'assumed similarity.’ Ppychological Bulletin, 1955, p2, No. 3: 177-193. Cronbach, L. J., and Glesser, G. C. Assessing similarity between pro- files. Psycholpgical Bulletin, 1953, pp, No. 6: 456-473. Crow, W. J., and Hammond, K. R. The generality of accuracy and response in interpersonal perception. Journal of Abnormal and Social Psychol- pgy, 1957, 54; 384-390. D'Andrade, R. G. Trait psychology and componential analysis. In E. A. Hammel (ed.), Formal American Semantic Anal Sis. Anthropologist Special Issue, 5, ___ , t. 2 : - Danes, J., and Woelfel, J. An Alternative to the Traditional Scaling Paradigm in Mass Communication Research: Multidimensional Reduction of Ratio Judgments of Separation. Paper presented at the annual meeting of the International Communication Association, Chicago, Illinois, April 1975. Doherty, L. M. The Relationship of Multidimensional Scaling Spaces of Trait Adjectives for Different Reference Persons. Unpublished doctor- al dissertation, Department of Psychology, University of Southern California, 1973. Einstein, A. Relativity. New York: Crown Publishers, 1961. Friendly, M. L., and Glucksberg, 5. On the description of subcultural lexicons: a MOS approach. Journal of Personalityand Social Psychol- pgy, 1970, 14; 55-65. 132 Gillham, J., and Woelfel, J. The Galileo System of Measurement: Prelim- inary Evidence for Precision, Stability and Equivalence to Traditional Measures. Unpublished paper, Department of Communication, Michigan State University, 1975. Gordon, T. F. Subject Abilities to Use Metric MDS: Effects of Varying the Criterion Pair. Paper presented at International Communication Association Convention, Portland, Oregon, April 1976. Gulliksen, H. How to make meaning more meaningful. Contemporary Psychology, May 1958, III, 5: 115-119. Green, R. F., and Goldfried, M. R. 0n the bipolarity of semantic space. Psychological Monogrpphs, 1965, 79, 6, #599: 1-131. Gross, C. F. Intrajudge consistency in ratings of heterogeneous others. Journal of Abnormal and Social ngchology, 1960: 605-610. Hanno, M. S., and Jones, L. E. Effects of a change in reference person on the multidimensional structure and evaluations of trait adjectives. Journal of Personality and Social Psychology, 1973, 28, 3: 368-375. Hastorf, A. H., Schneider, 0. J., and Polekfa, J. Person Perception. Menlo Park, California: Addison-Wesley, 1970. Hastorf, A. H., Richardson, 5. A., and Dornbush, S. M. The problem of relevance in the study of person perception. In R. Taguiri and L. Petrullo (eds.), Person Perception and Interpersonal Behavior, Stanford: Stanford University Press, 1958, pp. 54-62. Hays, W. L. An approach to the study of trait implication and trait similarity. In R. Taguiri and L. Petrullo (eds.), Person Perception and Interpersonal Behavior, Stanford: Stanford University Press, 1958: pp. 289-299. Heston, J. Ideal Source Credibility: A Re-examination of the Semantic Differential. Paper presented at International Communication Associa- tion convention, Montreal, Canada, April 1973. Horn, J. L. Factor analyses with variables of different metric. Educational and Psycholpgical Measurement, 1969, 29; 753-762. Hovland, C., Janis, F., and Kelley, H. Communication and Persuasion. New Haven: Yale University Press, 1958. Jaccard, J. J., and Fishbein, M. Inferential beliefs and order effects in personality impression formation. Journal of Personality and Social P§ychology, 1975, pl, 6: 1031-1040. Jackson, 0. N. The measurement of perceived personality trait relation- ships. In N. F. Washburne (ed.), Decisions, Values and Groups, Vol. 2, New York: Pergamon Press, 1962, pp. 177-188. \ /. I .1liuluw 133 Jones, E. E., and Hisbett, R. E. The actor and the observer: divergent perceptions of the causes of behavior. In E. E. Jones et al. (eds.), Attribution, Morristown, N. J.: General Learning Pressj'l971. Kaiser, H. F. The varimax criterion for analytic rotation in factor analysis. Psychometrika, 1958, 22; 187-200. Kelly, H. H. The warm-cold variable in first impressions of persons. Journal of Personality, 1951, 12; 431-439. Kelly, G. A. The Psycholpgy of Personal Constructs. New York: Norton, 1955. Koltuv, 8. Some characteristics of intrajudge trait intercorrelations. Psychological Monographs, 1962, Z§_(33, Whole No. 552): 1-33. n/Kruskal, J. B. Multidimensional scaling by optimizing goodness of fit to nonmetric hypothesis. Psychometrika, 1964a, 22; 1-27. ’Kruskal, J. B. Nonmetric multidimensional scaling: a numerical method. Psychometrika, l964b, 22; 115-129. Lay, C. H. Trait-inferential relationships and judgments about the per- sonalities of others. Canadian Journal of Behavioral Sciences, 1970, 2, 1: 1-17. Lingoes, J. C. A general survey of the Guttman-Lingoes nonmetric program series. In R. N. Shepard, A. K. Romney and S. B. Nerlove (eds.), J'Multidimensional Scaling; Theory and Application in the Behavioral Sciences, VOT} I, Theory, New Y0rk: Seminar Press, 1972. McCroskey, J. C. Scales for the measurement of ethos. Speech Monographs, 1966, 22; 65-72. McCroskey, J., Jensen, T., and Todd, C. The Generalizability of Source Credibility Scales for Public Figures. Paper presented at Speech Communication Association convention, Chicago, Illinois, December 1972. McCroskey, J., Jensen, T., Todd, C., and Toomb, J. K. Measurement of the Credibility of Organization Sources. Paper presented at International Communication Association convention, Montreal, Quebec, April 1973. McCroskey, J., Jensen, T., and Valencia, C. Measurement of the Credibil- ity of Peers and Spouses. Paper presented at International Communica- tion Association convention, Montreal, Quebec, April 1973. McLaughlin, M. L. Recovering the structure of credibility judgments: an alternative to factor analysis. Speech Monographs, 1975, 52; 221-228. Marlier, J. Procedures for a Precise Test of Social Judgment Predictions of Assimilation and Contrast. Paper presented at the Speech Communi- cation Association convention, Chicago, December 1974. 134 Messick, S. J. Metric properties of the semantic differential. Educa- tional and Psychological Measurement, 1957, 12; 200-206. Miller, G. A. A psychological method to investigate verbal concepts. Journal of Mathematical Psychology, 1969, p; 169-191. Mischel, W. Personalipy and Assessment. New York: Wiley, 1968. Mosteller, F., and Tukey, J. W. Data analysis, including statistics. In G. Lindzey and E. Aronson (eds.), Handbook of Social Psycholggy, Vol. 2, Reading, Massachusetts: Addison-weSTey,il969. Muliak, S. A. Are personality factor's raters conceptual factors? Journal of Consulting Psychology, 1964, 22; 506-511. Newcomb, T. An experiment designed to test the validity of a rating scale. Journal of Educational Psychology, 1931, 22; 279-289. Norman, W. T. Toward an adequate taxonomy of personality attributes. Journal of Abnormal and Social Psychology, 1963: 574-583. Osgood, C. E., Suci, G. J., and Tannenbaum, P. H. The Measurement of Meaning. Urbana, Illinois: University of IllinoisTPress, 19572 Passini, F. T., and Norman, W. T. A universal conception of personality structure? Journal of Personality and Social Psychology, 1966, 4: 44-49. T’ Pederson, D. M. The measurement of individual differences in perceived personality-trait relationships and their relation to certain deter- minants. Journal of Social Psychology, 1965, pp; 233-258. Phillips, J. The theory of configurational consistency. Unpublished manuscript, East Lansing, Michigan, Michigan State University, in progress. Rogers, E. M., and Shoemaker, F. F. Communication of Innovations. New York: Free Press, 1971. Rosenbaum, L. L., Rosenbaum, W. B., and McGinnies, E. Semantic differ- ential factor structure stability across subject, concept and time difference. Multivariate Behavioral Research, October 1971: 451-469. Rosenberg, S., Nelson, C., and Vivekananthan, P. S. A multidimensional approach to the structure of personality impressions. 'Journal of Personality and Social Psycholggy, 1968, p; 283-294. Rosenberg, S., and Olshan, K. Evaluative and descriptive aspects in personality perception. Journal of Personality and SocialPsycholpgy, 1970, 1g; 619-626. 135 Rosenberg, S., and Sedlak, A. Structural representations of perceived personality trait relationships. In R. N. Shepard, A. K. Romney, and S. Nerlove (eds.), Multidimensional Scalingz‘Theogy and Applications in the Behavioral Sciences, V01. II, New York: *Sem nar Press, 1972, pp. 1331162. Schneider, 0. J. Implicit personality theory: a review. Psychological Bulletin, 1973, 22, 5: 294-309. Scott, W. A. Structure of natural cognitions. Journal of Personality and Social Psychology, 1969, 12, 6: 294-309. Secord, P. F., and Berscheid, E. S. Stereotyping and the generality of implicit personality theory. Journal of Personality, 1962, 21; 65-78. Serota, K. 8. Metric Multidimensional Scaling and Communication: Theory and Implementation. M. A. Thesis, Department of Communication, Michigan State University, 1974. Shepard, R. N. The analysis of proximities: multidimensional scaling with an unknown distance function, I. Psychometrika, 1962a, 21, 2: 125-140. Shepard, R. N. The analysis of proximities: multidimensional scaling with an unknown distance function, II. Psychometrika, l962b, 27, 3: 219-246. _-' Shepard, R. N. Metric structures in ordinal data. Journal of Mathemat- ical Psychology, 1966, 2; 287-315. Shepard, R. N. A taxonomy of some principal types of data and of multi- dimensional methods for their analysis. In R. N. Shepard, A. K. Romney, and S. B. Nerlove, Multidimensional Scaling; Theory and Applications in the BehavioraT’Sciences,Vol. I, New Y6rk: Seminar Press, 1972. Sherman, R. Individual differences in perceived trait relationships as a function of dimensional salience. Multivariate Behavioral Research, 1972, 2; 109-129. Sherman, R. C., and Gillen, H. B. The Effects of Target Sex and Attrac- tiveness or Trait Attributions. Paper presented at the 83rd Annual American Psychology Association convention, Chicago, 1975. Shikiar, R., Fishbein, M., and Wiggins, N. Individual differences in semantic space: a replication and extension. Multivariate Behavioral Research, Apri1, 1974: 200-209. Shrauger, J. S., and Patterson, M. B. Self-evaluation and the selection of dimensions for evaluating others. Journal of Personality, 1974, 42, 4: 569-585. Snider, J. G., and Osgood, C. E. Semantic Differential Technique. New York: Aldine-Atherton, 1969. 136 Snyder, F. W., and Wiggins, N. Affective meaning systems: a multivar- iate approach. Multivariate BehaviOral'Research, October, 1970: 453-468. Talbott, A. D. A Study of the Generality of Dimensions of Source Evalu- ation. Unpublished Dissertation, Department of Communication, Michigan State University, 1968. Tannenbaum, P. Mediated generalization of attitude change via the prin- ciple of congruity. Journal of Personality and Social Psycholggy, 1966: 493-499. Tannenbaum, P. The congruity principle: retrOSpective reflections and recent research. In Abelson et pl, (eds.), Theories of Cognitive Consistency: A Sourcebook, Chicago: Rand Me a y, , pp. - 2. Thorndike, E. L. A constant error in psychological rating. Journal of Applied Psyghology, 1920, 3; 25-29. Thurstone, L. L. A law of comparative judgment. Psychological Review, 1927, 25; 273-286. Thurstone, L. L. Multiple Factor Analysis. Chicago: University of Chicago Press, 1947. Todd, F. S., and Rappoport, L. A. A cognitive structure approach to person perception: a comparison of two models. Journal of Personal- ity and Social Psycholggy, 1964, pg, 5: 469-475. Torgerson, W. S. Theory and Methods of Scaling. New York: John Wiley and Sons, Inc., 1958. Tucker, L. R., and Messick, S. An individual differences model for mul- tidimensional scaling. P§ychometrika, 1963, 22; 333-367. Veldman, D. J. Simple structure and the number of factors problem. Multivariate Behavioral Research, April, 1974: 191-201. Walters, H. A., and Jackson, 0. N. Group and individual regularities in trait inference: a multidimensional scaling analysis. Multivariate Behavioral Research, 1966, l; 145-163. Warr, P. 8., and Simms, A. A. Study of co-judgment processes. Journal of Personality, 1965, 22; 598-604. Warr, P. 8., and Knapper, C. The Perception of People and Events. New York: Wiley, I968. Warr, P. 8., and Smith, J. S. Combining information about people: com- parisons between six models. Journal of Personality and Social Psy- chology, 1970, 1g, 1: 55-65. Whitehead, A. N. Science in the Modern World. New York: The Free Press, 1969. 137 Wiggins, N., and Fishbein, M. Dimensions of semantic space: a problem of individual differences. In J. R. Snider and C. E. Osgood (eds.), The Semantic Differential Technique, Chicago: Aldine, 1969, pp. 183- 1931 Wishner, J. Re-analysis of 'Impressions of Personality.’ ngchological Review, 1960, p]; 96-112. Wittgenstein, L. Philosophical Investigations. G. E. M. Anscombe and R. Rhees (eds.). G. E.1M. Anscombe (trans.). Osford: Blackwell Press, 1953. Woelfel, J. Metric Measurement of Cultural Processes. Paper presented to the Speech Communication Association, Chicago, Illinois, December 1974. Woelfel, J., Fink, E., Holmes, R., Cody, M., and Taylor, J. A Mathemat- ical Procedure for Optimizing Political Campaign Strategy. Paper presented at political division, International Communication Associa- tion convention, Portland, April 1976. Woelfel, J., Saltiel, J., McPhee, R., Serota, K., Barnett, G., Danes, J., and Cody, M. Orthogonal Rotation to Theoretical Criteria: Comparison of Multidimensional Spaces. Paper presented at the annual convention of the Mathematical Psychology Association, Purdue, Indiana, 1975. Young, F. W., and Torgerson, W. S. TORSCA, a Fortran IV program for Shepard-Kruskal multidimensional scaling analysis. Behavioral Science, 1967, l2; 498. Zajonc, R. 8. Cognitive theories in social psychology. In G. Lindzey and E. Aronson (eds.), Handbook of Social Psychology, Vol. 1, Reading, Massachusetts: Addison-Wesley Press, 1969. APPENDIX MICHIGAN STATE UNIVERSITY COLLEGE OF COMMUNICATION AR'I'S EAST IANSING ° MICHIGAN ° 48824 DEPAR'I'MI'IN'I‘ OF COMMUNICATION Dear Participant: The research that you will participate in today is very important to me and I would first like to thank you for volunteering and taking your time to come in. I am sure that you are first interested in the goal of this research task. Easentially, it is concerned with language; and more specifically with those adjectives peOple use in either describing themselves to others or in their descriptions of others. Each adjective is a guality that a person can possess. What we are interested in today is to see M m 2592!. m qualities lp_terms gg_how desirable they are lg_yeur idea 9; §_close 251222. This will become clearer when we present the instructions in a minute. First, let me say that if you are interested in how this research corresponds with other research that we are doing, please see we at your convenience; if you are interested in the results, please see he in class next week. Finally, let me thank you again for caning in today. Thank you, \. J .. , ((1- Michael Cody MICHIGAN STATE UNIVERSITY COLLEGE OF COMMUNICATION ARTS EAST LANSING ' MICHIGAN ° 48824 DEPARTMENT OF COMMUNICATION INSTRUCTIONS Before you there is a deck of cards. You will find 993 word typed on each card. These words are adjectives typically used by peeple to describe qualities that peeple possess. Your task is to sort these qualities into categories. On the table in front of you, you will find ELEVEN spaces in which you may place cards. Each space is a category. On the far lei} category, place those qualities that fit the following criteria: THIS QUALITY IS A QUALITY THAT A CLOSE FRIED OF MINE m POSSESS. On the far M13 category, place those qualities that fit the following criteria: THIS QUALITY IS A QUALITY THAT A CLOSE FRIEND OF MINE SHOJLD Egg: Possess. USE THE OTHER CATEGORIES TO PLACE THOSE QUALITIES WHICH ARE OF LIES IMPORTANCE FOR A CLOSE FRIEND. FOR EXAMPLE, IN THE FAR LEFT CATEORY, PLACE THE QUALITIES THAT A FRIEND §1;I_O_1_J_LQ POSSESS, AND PLACE THE QUALITIES YOU w _L_I_K_E_ A FRIEND TO HAVE IN THE CATEORY TO THE RIGHT OF IT. SO, QUALITIES ARE TO BE PLACED _IN A DESCENDING ORDER QF_ HOW DESIRABLE THEY ARE FROM THE FAR £13111: CATEORY 19 1112 [All 32953 CATmORY. Are there any questions? Please take a few minutes and read the qualities typed on the cards: then begin. Take as long as you wish. Thank you. Positive induction message: Birch Bayh has demonstrated his skill and proficiency in public service time and time again. He is competent, demonstrably capable and decidedly qualified to address the issues in America today. He is a seasoned veteran whose background and past experiences in public life have made him one of the most experienced men on the political scene to- day. Many analysts have commented on the similarities between Birch Bayh and Hubert Humphrey. Indeed, Birch Bayh may be viewed as the Humphrey legatee; an advocate of that which Humphrey represents—35d has represented since he entered politics. Their philosophies are similar. Their experiences and personalities are similar. In addition, Birch Bayh deals fairly with issues. He is just, even-handed and un-biased in his speeches. Negative induction message: Birch Bayh has demonstrated his lack of skill and proficiency in public service time and time again. He is incompetent, demonstrably in- capable and decidedly unqualified to address the issues in America today. He is a political rookie whose background and lack of experiences in public life have made him one of the least experienced men on the polit- ical scene today. Many analysts have commented on the contrasts between Birch Bayh and Hubert Humphrey. Indeed, Birch Bayh may be viewed as the antithesis of Humphrey; the opposite of everything that Humphrey repre- sents and has represented since he entered politics. Their philosophies are different. Their practice of politics is different. Their experi- ences and personalities are different. In addition, Birch Bayh deals unfairly with issues. He is unjust, one-sided and biased in his speeches. Dear Participant: In the Department of Communication, we are engaged in ongoing research concern- ing student attitudes on a variety of topics. The purpose of this survey is to gauge your attitude on several political and social issues. INSTRUCTIONS Below you will find several statements. Each statement expresses an attitude toward some topic. Please.indioate the extent to which you AGREE or.DISAGREE with the attitude expressed in the statement. For example, if you strongly agree with the attitude expressed in the statement, you would place an X in the space nearest to "AGREE," as shown below. AGREE / X / / / / / / / / / / / DISAGREE Note that there are eleven spaces. Space (1) represents veg! gtrgng_gg£3ement with the attitude expressed in the statement, and space (11) on the far right, closest to DISAGREE, represents very strong disagreement with the attitude ex- pressed in the statement. Similarly, if you are in only slight agreement with the attitude expressed in the statement, use one of the spaces to the right of the first space. If you neither agree or disagree with the statement, or have no opinion on the subject, then place your X in the middle space; space (6). If there are any questions, please ask. Thank you in advance for your coopera- tion with this project. Michael Cody In order to sort questionnaires, please write the last four numbers of your student I.D.: Today's Date Statement.l: IT IS THE OBLIGATION OF THE FEDERAL GOVERNMENT TO ENABLE ALL AMERICAN CITIZENS ACCESS TO DECENT HEALTH CARE THROUGH A SOCIALIZED MEDICINE PROGRAM. AGREE / / / / / / / / / / / / DISAGREB l 2 3 u 5 6 7 8 9 10 11 Statement 2: CONGRESS SHOULD DRASTICALLY CUT BACK ON THE INFLATED PENTAGON BUDGET. msacm/ //// /___//////AGREE Statement 3: SCHOOL BUSING TO ACHIEVE RACIAL BALANCE IN SCHOOLS HILL WORSEN RACIAL RELATIONS-- NOT IMPROVE 0N THEM. AGREE / / / / / / / / / / / / DISAGREE Statement u: ‘WITH THE RAPID DEVELOPMENT AND PRODUCTION OF NUCLEAR ARMS, SUPPORT FOR DETENTE IS CRUCIAL AND IMPERATIVE. nxsacnsn / / / / / /__/__/____/ / / / acme: l 2 3 u 5 6 7 8 9 10 ll Statement 5: LEGISLATION SHOULD BE ENACTED IMMEDIATELY TO CURTAIL THE OIL COMPANIES' GIGANTIC PROFITS AT THE CONSUMER’S EXPENSE. AGREE / / / / / / /___/ / / /___/ nxsmm l 2 3 H S 6 7 8 9 IO 11 Statement 6: THE UNITED STATES SHOULD WITHDRAW ALL SUPPORT, BOTH ARMS AND FINANCIAL, FROM THE CONFLICT IN ANGOLA. DISAGREB / / / / / / / / / / / / AGREE l 2 3 M 5 6 7 8 9 10 ll Statement 7: THE TIME IS LONG OVERDUE T0 GRANT AN UNCONDITIONAL AMNESTY TO AMERICAN MEN STILL IN EXILE IN FOREIGN COUNTRIES BECAUSE OF THEIR STAND AGAINST THE VIETNAM WAR. AGREE / / / / /____/ / / / /___/___/ DISAGREE l 2 3 H 5 6 7 8 9 10 ll Statement 8: EDUCATION IS VALUABLE FOR ITS OWN SAKE, EVEN IF IT DOESN'T PREPARE YOU FOR A JOB. DISAGREE////////////AGREE 1 2 3 H 5 6 7 8 9 10 ll Statement 9: THE SALE AND USE OF MARIJUANA SHOULD BE LEGALIZED. AGREE / / / / /____/__/___/ / / / / DISAGREE l 2 3 u 5 6 7 8 9 10 11 Statement 10: FEDERAL SPENDING OUGHT TO BE CURTAILED IN ORDER TO BRING THE NATIONAL BUDGET INTO BALANCE. DISAGREE / / / / / / / / / /____/___/ AGREE 12314567891011 Statement 11: THE FEDERAL GOVERNMENT SHOULD TAKE OVER.OWNERSHIP AND OPERATION OF ALL RAILROADS IN THE UNITED STATES. AGREE////////////DISAGREE 123u567891o.11 Statement 12: ANY LEVEL OF UNEMPLOYMENT CAN BE TOLERATED FOR A FEW YEARS IF IT WILL HELP TO END INFLATION. DISAGREE / / / / / / / / / / / / AGREE l 2 3 H 5 6 7 8 9 10 11 Finally, we would like some information about you. There are several questions below that we would like you to answer. If, for any reason, you do not want to answer one or more of these questions, please feel free to skip it. How- ever, try to answer as many questions as possible. Let me remind you that this information is kept strictly confidential. It is needed only to compare samples of subjects who are filling out this questionnaire and other question- naires similar to it. Again, thank you for your time. 1. What is your age? - 9 2. What 18 your sex. Male (Circle one) Female 3. What is your year in school? Freshman Sophomore Junior (Circle one) Senior Graduate student u. What is your major? (If no major, please write "none") 5. Race. American Indian Black Caucasan Chicano Oriental Other (please specify) (Circle one) Dear Participant: In the Department of Communication we are engaged in ongoing research con~ cerning student attitudes on a variety of topics. Because familiarity with an object, concept or person is an important component of a person's attitude, this particular study seeks to discover student familiarity with a variety of persons in American politics. In the attached questionnaire we would like you to answer six questions con- cerning each person presented. In the first question, we would like you to give us your estimate of YOUR FANILIARITY with the person. If you are very familiar with who this person is, check the far LEFT of the scale. For example: very familiar / X / / / / / not familiar 1 2 3 H 5 If this person is not familiar to you, then check the far RIGHT. Question 2 asks the question, 'Thiguaerson is extremely important in American politics." If you believe that this person is extremely important in American politics, then respond by circling the far LEFT statement. For example: GEEE§\» Somewhat Somewhat Strongly Neutral . . Agree Disagree Disagree Str Agree If you believe that this person is not important in American politics, then cirei' the far RIGHT end of the scale. The next two questions are very straightforward. They ask: "What is this person's occupation?" and "What notable things has this person done?" Questions five and six ask how many times you have heard about this person from family and friends in the last two weeks, and how many times you have heard about this person from the media (newspapers, T.V., radio, and magazines) in the past two weeks. If you have not heard anything from your family or friends about this person in the last two wEEks, then write a zero (0) in the space provided. If you have heard something about this person in the last two weeks, try to esti~ mate how many times you have heard him or her discussed. Use any number you feel is appropriate in the space provided. If you have any questions at this point, please raise your hand. Be assured that your responses will be kept in the strictest confidence. Your cooperation and careful consideration on each question will be greatly appreciated. Thank you. Michael Cody The 1. The following questions are about Edmond (Jerry) Brown. How familiar are you with Edmond (Jerry) Brown? Very familiar / / / / / / Hot familiar l 2 3 4 5 Edmond (Jerry) Brown is extremely important in American politics. (circle one) Strongly Somewhat . Somewhat Strongly Neutral . . Agree Agree Disagree Disagree What is Edmond (Jerry) Brown's occupation? What notable things has this person done? In the past two weeks, how many times have you heard from the media (newspa- pers, T.V., radio, magazines, etc.) about Edmond (Jerry) Brown? In the past two weeks, how many times have you heard from family, friends, and acquaintances about Edmond (Jerry) Brown? :‘:*:':**:’: following questions are about Jimmy Carter. How familiar are you with Jimmy Carter? Very familiar / / / / / / Not familiar 1 2 3 E 5 Jimmy Carter is extremely important in American politics. (circle one) Strongly Somewhat Neutral Somewhat Strongly Agree Agree Disagree Disagree What is Jimmy Carter's occupation? What notable things has this person done? In the past two weeks, how many times have you heard from the media (newspa- pers, T.V., radio, magazines, etc.) about Jimmy Carter? In the past two weeks, how many times have you heard from family, friends, and acquaintances about Jimmy Carter? K7 The 1. following questions are about Nelson Rockefeller. How familiar are you with Nelson Rockefeller? Very familiar / / / / / / Not familiar l 2 3 u 5 Nelson Rockefeller is extremely important in American politics. (circle one) Strongly Somewhat Neutral Somewhat Strongly Agree Agree “ ‘ Disagree Disagree What is Nelson Rockefeller's occupation? What notable things has this person done? In the past two weeks, how many times have you heard from the media (newspa- pers, T.V., radio, magazines, etc.) about Nelson Rockefeller? In the past two weeks, how many times have you heard from family, friends, and acquaintances about Nelson Rockefeller? AAAAAA following questions are about Shirley Chisholm. How familiar are you with Shirley Chisholm? Very familiar / / / / / i Not familiar l 2 3 u 3 Shirley Chisholm is extremely important in American politics. (Circle one) Strongly Somewhat . Somewhat Strongly Neutral . . Agree Agree Disagree Disagree What is Shirley Chisholm's occupation? What notable things has this person done? In the past two weeks, how many times have you heard from the media (newspa- pers, T.V., radio, magazines, etc.) about Shirley Chisholm? In the past two weeks, how many times have you heard from family, friends, and acquaintances about Shirley Chisholm? The 10 following questions are about Hubert H. Humphrey. How familiar are you with Hubert H. Humphrey? Very familiar / / / / / / Not familiar l 2 3 H 5 Hubert H. Humphrey is extremely important in American politics. (circle one) Strongly Somewhat , Somewhat Strongly Neutral . . Agree Agree Disagree Disagree What is Hubert H. Humphrey's occupation? What notable things has this person done? In the past two weeks, how many times have you heard from the media (newspa- pers, T.V., radio, magazines, etc.) about Hubert H. Humphrey? In the past two weeks, how many times have you heard from family, friends, and acquaintances about Hubert H. Humphrey? assess fbllowing questions are about George C. McGovern. How familiar are you with George C. McGovern? very familiar / / / / / / Not familiar l 2 3 u 5 George C. McGovern is extremely important in American politics. (circle one) Strongly Somewhat Neutral Somewhat Strongly Agree Agree Disagree Disagree What is George C. McGovern's occupation? What notable things has this person done? In the past two weeks, how many times have you heard from the media (newspa- pers, T.V., radio, magazines, etc.) about George C. McGovern? In the past two weeks, how many times have you heard from family, friends, and acquaintances about George C. McGovern? O The following questions are about Bella Abzug. 1. How familiar are you with Bella Abzug? Very familiar / / / / / / Not familiar l 2 3 u 5 Bella Abzug is extremely important in American politics. (circle one) Strongly Somewhat . Somewhat Strongly Neutral . . Agree Agree Disagree Disagree What is Bella Abzug's occupation? What notable things has this person done? In the past two weeks, how many times have you heard from the media (newspa- pers, T.V., radio, magazines, etc.) about Bella Abzug? In the past two weeks, how many times have you heard from family, friends, and acquaintances about Bella Abzug? :‘efcfiéfif: MIGiIGON STATElNIVEISITY COLLEGE OF COMMUNICATION ARTS DEPARTMENT op COMMUNICATION EAST LANSING, MICHIGAN, neezu Dear Participant: In the Department of Communication, we are engaged in ongoing research concerning student attitudes on a variety of topics. The purpose of this questionnaire is to see how you perceive others. Specifically, we are interested in how you perceive several political figures. We would like to know how you perceive these people in relation to words typically used to describe political figures; i.e., trait adjectives such as "informed", "experienced", etc. Finally, we are interested in how you perceive the concept of the "Ideal Credible Source". Imagine that the "Ideal Credible Source" is an individual who on LIKE TO HEAR SPEAK,;whose opinions YOU TRUST 294 WHOSE ADVICE YOU HOUED LISTEN TO. To give you some information which will be helpful to you in filling out the questionnaire we have included, on the next pages, a descrip- tion of a public figure and some excerpts from a recent speech. This will give you an idea of what the candidate is like and what he advo- cates. Please read this information carefully and thoroughly, it will be helpful to you. Please read the instructions carefully. If you have any questions, please ask the researcher. I will gladly discuss our research with you at any time. All answers will be kept in strict confidence. However, in order to sort the different questionnaires and to keep the questionnaires separate, please write the last four numbers of your student I.D.: Thank you, Michael Cody I C N {2 “The following description of Birch Bayh is taken from a pamphlet written and <3irculated by "Nonpartisan Citizens Committee for the Promotion of Informed ‘Voting". This group is composed largely of independents and is active pri- rmarily in Minneapolis and Indianapolis. The group has analyzed a large number of nationally known political figures in terms of voting record, interests and qualifications. The following is what the group had to say about Birch Bayh: "Birch Bayh has demonstrated his skill and proficiency in public service time and time again. He is competent, demonstrably capable and decidedly qualified to address the issues in America today. He is a seasoned veteran whose back- ground and past experiences in public life have made him one of the most ex- perienced men on the political scene today. Many analysts have commented on the similarities between Birch Bayh and Hubert Humphrey. Indeed, Birch Bayh may be viewed as the Humphrey legatee; an advocate of that which Humphrey represents and has represented since he entered politics. Their philosophies are similar. Their practice of politics is similar. Their experiences and personalities are similar. In addition, Birch Bayh deals fairly with issues. He is just, even handed and unbiased in his speeches. " The following are excerpts from a speech recently given by Birch Bayh: Nothing is more important to the American electorate than inflation and unemployment. Americans are concerned about their neighbors who are out~ of work and forced to struggle to get by. Americans are concerned about the rising cost of living, which has ferced everyone to skimp and struggle. This nation cannot endure a painful "go slow" prescription fer economic recovery. ’ We cannot afford to cut back on Government sponsored child-care centers and food programs. We cannot afford to cut back on Government sponsored job training programs for the disadvantaged. We cannot afford to cut back on jobs. Jobs are the key to getting this nation moving again. We need to increase employment and to‘ggt_money_ to circulate more freely_ in the economy. We cannot afford to spend time now quibbling about the federal government's deficit. We cannot sacrifice our nation' 8 recovery_b reducing government 8 endin . We must increase Soc1al Security benefits or t e elderly whose fgxed incomes have unjustly been strangled by inflation. We must provide more jobs to pe0ple---to stimulate this economy, get people off the streets and back on the job. A "go slow" prescription for economic recovery calls for balancing the budget. Doing 23 will keep unemployment 32 this nation way over Z_percent. The last time the federal government reduced spending in order to reduce the ‘ federal deficit it was at the undue expense of the poor, the handicapped and the elderly on fixed incomes. In addition, it put undue pressures on young people in need of scholarship aid and training programs for the needy thus preventing them a fair start on the road to becoming successful and produc- tive citizens. It is in our best interest not to cut back on these programs. Justice and decenCy demand that we continue to provide at least the minimum necessities for a decent life to those among us who, through no fault of their own, cannot provide for themselves. A reasoned and prudent concern for our own futures and those of our children dictates that we continue to invest in the development of our most valuable national resource, our young people. If a balanced budget means surrendering our humanitarian concern for those less fortunate, and if it means surrendering our prudent concern for pre- paring our children to keep America great in the future--and it does mean surrenderi_g_those concerns--then we cannot afford it at this time. We wOuld like you to respond to a few additional attitude statements. Recall that space (1) represents very stron agreement with the attitude expressed in the statement, and space 111 on the far right, closest to DISAGREE, rep— resents very strong disagreement with the attitude expressed in the statement. Similarly, 1f you are in only slight agreement with the attitude expressed in the statement, use one of the spaces to the right of the first space. If you neither agree or disagree with the statement, or have no opinion on the sub- ject, then place your X in the middle.space; space (6). Statement 1: FEDERAL SPENDING OUGHT TO BE CURTAILED IN ORDER TO BRING THE NATIONAL BUDGET INTO BALANCE. ' AGREE / / / / / / / / /‘ / / / DISAGREE 1 2 3 4 5 6 7 8 9 10 ll Statement 2: CANDIDATES WHO RUN FOR PUBLIC OFFICE SHOULD AGREE WHEN THEY DO SO TO SURRENDER THEIR RIGHT TO.PRIVACY AND EXPOSE ALL OF THEIR AFFAIRS TO PUBLIC SCRUTINY. AGREE / / / / / / / '/ / / / / DISAGREE l 2 3 u 5 6 7 . 8 9 10 11 - Statement 3: THE GOVERNMENT SHOULD SPEND AS MUCH MONEY AS NECESSARY IN ORDER TO START FEDERAL PROGRAMS TO PROVIDE JOBS FOR THE UNEMPLOYED WHO WANT TO WORK. AGREE /__/___l / J / / / / / / / DISAGRBB 123u567891011 INSTRUCTIONS In this questionnaire, we would like you to tell us how different (or, in other works "how far apart") certain people and trait adjectives are from each other. Differences between these concepts can be measured in Social Units. "Social Units" refer to the psychological distance or dissimilarity between people and words; The meaning of some words may be extremely different or they may be extremely similar. To aid in understanding what we mean by "distance", please use the following rule: '- The distance between INTELLIGENT and _INEXPERIENCED 33 100 Social Units. We would like you to tell us how many Social Units apart the people and trait adjectives are from each other. Remember, the more different they are from each other, the bigger the number of Social Units- The less different they are, the smaller the number of Social Units. The following can serve an an example. First you are given the rule and then you are asked to report the distances between pairs of words and_persons. For example: IF THE DISTANCE BETWEEN INTELLIGENT AND INEXPERIENCED IS 100 UNITS, HON FAR APART ARE? Mickey Mouse and awful units Mickey Mouse and nice units awful and nice units If you perceive Mickey Mouse as being glg§g_to the meaning you have for the word "awful", then you would write, in the space provided, a small number of Social Units. If you perceive Mickey Mouse as very distant from "awful", then you would write a number that is very large. Similarly, if you perceive MiCkey House as close to "nice”, report a small .nUmber of Social Units in the space provided, and if you perceive that Mickey Mouse is very far from "nice", then report a larger number of Social Units. FEEL FREE TO USE ANY NUMBER TO REPORT AS ACCURATELY AS POSSIBLE THE DISTANCE THAT YOU SEE BETWEEN THE TWO CONCEPTS. You may use a number over 100 if the distance between any two concepts is greater than the distance between the concepts INTELLIGENT and INEXPERIENCED. If you perceive two words to be extremely similar to each other in meaning, then report a very small number. If you think that there is no Ci’fcrencc in +3~ wq-fi4ng of two concepts, then you may write zero (0) to represent no distance between them. 0n the following pages, you will find lists of pairs of words similar to those shown above. Please write a-number in the blank space after each pair of concepts. Ignore the column of numbers next to the blanks; they are for clerical use only. Please try not to skip any item. Try to report some distance between each pair of concepts. Keep in mind that there is no one correct answer; all that we ask is that you give us honest and careful responses about how you perceive the meaning of these trait adjectives and how these adjectives relate to these people. If you have any questions, feel free to ask the researcher. Thank you. REMEMBERI IF INTELLIGENT AND INEXPERIENCED ARE lOO UNITS APART ARE: APART, How FAR ID# l-u Group 5 Wave 6 How far apart are: Card 01 7-8 competent and inexperienced units 9-13 competent and Birch Payh .... units 1n-13 competent and repulsive :::: units 19-23 competent and unintelligent 'rnits 2u-29 competent and just __"muunits ' 29-33 competent and reliable units 3u-33 competent and unreliable -.units 39-“3 competent and Hubert Humphrey units uu—ua competent and unjust ‘ ___, units #9-53 competent and intelligent units SH-SB competent and Ideal Credible Source units 59-63 competent and attractive _::@nits 6u-68 competent and incompetent units 69-73 competent and experienced units 7ue78 DUP 1-6 How far apart are: Card 02 7-8 competent and George McGovern units 9-13 inexperienced and Birch Bayh -_—"—units lu-13 inexperienced and repulsive units 19-23 inexperienced and unintelligent units 2u-23 inexperienced and just units 29-33 inexperienced and reliable units 3u-38 inexperienced and unreliable ___——units 39-“3 inexperienced and Hubert Humphrey units uu-u8 inexperienced and unjust ' _—”‘”units u9-53 inexperienced and intelligent units su-58 inexperienced and Ideal Credible Source units 59-63 inexperienced and attractive units 6u_68 inexperienced and incompetent units 69-73 inexperienced and experienced units 7u-78 DUP 1-6 How far apart are: Card 03 7-8' incrferirnccd and George McGovern units 9-13 Birch Bayh and repulsive units 1n-13 Birch Bayh and unintelligent ‘units 19-23 Birch Bayh and just units 29-28 Birch Bayh and reliable units 29-33 Birch Bayh and unreliable units 3n-33 Birch Bayh and Hubert Humphrey ‘units 39-u3 Birch Bayh and unjust ___units un-u3 Birch Bayh and intelligent units u9-53 Birch Bayh and Idea? irclible Source -”'unifs ’ 54-58 Birch Bayh and attractive ~ units 59-63 Birch Bayh and incompetent units su—ee Birch Bayh and experienced units 69-73 Birch Bayh and George McGovern units 7u—78 Aa IF INTELLIGENT AND INEXPERIENCED ARE FAR APART ARE: REMEMBER: How far apart are: repulsive and unintelligent units repulsive and just units repulsive and reliable 'units repulsive and unreliable units repulsive and Hubert Humphrey units repulsive and unjust units repulsive and intelligent units 'repulsive and Ideal Credible Source units repulsive and attractive units repulsive and incompetent units repulsive and experienced units repulsive and George McGovern units unintelligent and just units unintelligent and reliable . units How far apart are: unintelligent and unreliable,_ units 'unintelligent and Hubert Humphrey units unintelligent and unjust units unintelligent and intelligent units unintelligent and Ideal Credible Source units unintelligent and attractive units unintelligent and incompetent units unintelligent and experienced units unintelligent and George McGovern units just and reliable units b just and unreliable units just and Hubert Humphrey units just and unjust units just and intelligent units How far apart are: just and Ideal Credible Source units just and attractive units just and incompetent units juSt and experienced unitS' just and George McGovern units reliable and unreliable units ‘ reliable and Hubert Humphrey units reliable and unjust units reliable and intelligent units reliable and Ideal Credible Source units reliable and attractive units reliable and incompetent units reliable and experienced units reliable and George Vchvern _"_—units ‘DUP 100 UNITS APART, HOW 1-6 Card on 7—8 DUP 9-13 lu-IB 19-23 2u-28 29—33 3M-38 39-“3 un-ue ”9-53 su-sa 59-63 6u~68 69—73 7H-78 1-6 Card 05 7—8 DUP 9-13 Ia-Ie 19—23 2u—2e 29-33 au—ae 39-u3 uu-u8 u9-53 su-58 ’59-63 SU-68 69-73 7u-73 1-6 Card 06 7-8 9-13 lu-lB 19-23 2u-2e 29-33 3M-38 39-“3 uu-ue #9-53 SQaSB 59-63 6u~68 69-73 7u-78 REMEMBERz. IT INTELLIGENT AND INEXPERIENCED APART ARE: 'How far apart are: unreliable and Hubert Humphrey units unreliable and unjust units unreliable and intelligent units unreliable and Ideal Credible Source units unreliable and attractive units unreliable and incompetent units unreliable and experienced units unreliable and George McGovern units Hubert Humphrey and unjust units Hubert Humphrey and intelligent units Hubert Humphrey and Ideal Credible Source units Hubert Humphrey and attractive units Hubert Humphrey and incompetent units Hubert Humphrey and experienced units How far apart are: ‘ Hubert Humphrey and George McGovern units unjust and intelligent units unjust and Ideal Credible Source units unjust and attractive units unjust and incompetent units unjust and experienced units unjust and George McGovern units intelligent and Ideal Credible Source units intelligent and attractive units intelligent and incompetent units intelligent and experienced units intelligent and Gacrge McGovern 7 units Ideal Credible Source and attractive. units Ideal Credible Source and incompetent units How far apart are: Ideal Credible Source and experienced units Ideal Credible Source and George.McGovern units attractive and incompetent units attractive and experienced units attractive and George McGovern units incompetent and experienced units incompetent and George McGovern units units experienced and George McGovern Ac DUP ARE 100 UNITS APART, HOW FAR 1-6 Card 07 7-8 DUP 9-13 lu-18 19-23 2H-28 29-33 3H-38 39-“3 flu-HB #9-53 SH—SB 59-63' su-ee 69-73 7H-78 1-6 Card 08 7-8 DUP 9-13 Iu-Ie 19-23 2u-2e. 29-33 au-3e ag-ua nu-ue u9_53 su—sa 59-63 >6u-68 69-73 7n—7a 1-6 Card 09 7-8 9-13 1u-18 19-23 2u—28 29-33 su-ae 39-u3 nu-ue Col. 'DUP. 1-3 RD £2.4-5 10 11 12 INSTRUCTIONS: Another pOpular way of making comparisons is indirectly, through the use of rating scales. In this section you will be asked to fill out a number of such scales. For example, you might be asked to rate the con- cept "weather." The scale would look like this: (WEATHER) GREAT: : : : : X : ' : :LOUSY 3 2 l 0 l 2 3 Strongly Applies Applies Sort of Applies Neutral of Doesn't Apply OHMOJ II II II II In this example, the person marked the spaCe indicating his feelings about the weather were "sort of lousy." Had the person marked the one (1) on the "GREAT" side of the "0" space, it would be an indication that the weather had been "sort of great." Please respond to the following scales according to your feelings. Herk rather quickly, as your initial response is probably the most accurate. Read each scale term carefully, however, before responding. Thank you. (HUBERT HUMPHREY) COMPETENT: : : : : : : :INCOMPETENT 3 2 l O l 2 3 Now, how IMPORTANT or UNIMPORTANT is the COMPETENT/INCOMPETENT distinction to your conception of HUBERT HUMPHREY? IMPORTANT: : ‘ : : : ’ : : :UNIMPORTANT 3 2 l O l 2 3 And, how RELEVANT or IRRELBVANT is the COMPETENT/INCOMPETENT distinction to your conception of HUBERT HUMPHREY? IRRELEVANT: : : : :‘ : : :RELEVANT 3 2 l 0 l 2 3 (HUBERT HUMPHREY) EXPERIENCED: : : : : : : :INEXPERIENCED 31 32 33 3Q 35 36 37 38. 39 Now, how IMPORTANT or UNIMPORTANT is the EXPERIENCED/INEXPERIENCED dis- tinction to your conception of BIRCH BAYH? UNIMPORTANT: :IMPORTANT 3 2 1 0 1 2 3 And, how RELEVANT or IRRELEVANT is the EXPERIENCED/INEXPERIENCED dis- tinction to your conception of BIRCH BAYH? RELEVANT: : : : : : : :IRRELEVANT 3 2 l 0 l 2 3 (BIRCH BAYH) ATTRACTIVE: : : : : : : :REPULSIVE 3 2 . l O l 2 ‘ 3 Now, how IMPORTANT or UNIMPORTANT is the ATTRACTIVE/REPULSIVE distinc- tion to your conception of BIRCH BAYH? IMPORTANT: : -:‘ : : ' : ' : :UNIMPORTANT And, how RELEVANT or IRRELEVANT is the ATTRACTIVE/REPULSIVE distinction to your conception of BIRCH BAYH? IRRELEVANT: : : : ‘ : : : :RELEVANT 3 2 p l 0 l 2 3 (BIRCH BAYH) INTELLIGENT: :UNINTELLIGENT 3‘ 2'1 0 1 2 3 Now, how IMPORTANT or UNIMPORTANT is the INTELLIGENT/UNINTELLIGENT dis- ‘ tinction to your conception of BIRCH BAYH? IMPORTANT: : : :UNINPORTANT 3 2 l o l 2 3 And, how RELEVANT or IRRELEVANT is the INTELLIGENT/UNINTELLIGENT dis- tinction to your conception of BIRCH BAYH? RELEVANT: :‘ : : : ' : : :IRRELEVANT 3 2 1 .o 1 2 3 (BIRCH BAYH) JUST: : : : : : : :UNJUST -10- 22 23 2M 25 26 27 28 29 30 Now, how IMPORTANT or UNIMPORTANT is the JUST/UNJUST distinction to your conception of HUBERT HUMPHREY? IMPORTANT: : : : : : : :UNIMPORTANT 3 2 1 O l 2 3 And, how RELEVANT or IRRELEVANT is the JUST/UNJUST distinction to your conception of HUBERT HUMPHREY? IRRELEVANT: : : : : : : ##5RELEVANT 3 2 1 O l 2 3 (HUBERT HUMPHREY) ‘ RELIABLE: : : : : : : :UNRELIABLE 3 2 1 _ o l 2 3 Now, how IMPORTANT or UNIMPORTANT is the RELIABLE/UNRELIABLE distinction to your conception of HUBERT HUMPHREY? IMPORTANT: : : : i : : : :UNIMPORTANT 3 2 l o - 1 2 3 And, how RELEVANT or IRRELEVANT is the RELIABLE/UNRELIABLE distinction to your conception of HUBERT HUMPHREY? IRRELEVANT: : : : : : : :RELEVANT (BIRCH BAYH) COMPETENT: : : : : : : ‘:INCOMPETENT 3 2 l 0 l 2 3 Now, how IMPORTANT or UNIMPORTANT is the COMPETENT/INCOMPETENT distanc- tion to your conception of BIRCH BAYH? IMPORTANT: : ": vfl_: : : : :UNIMPORTANT 3 2 l O ' l 2 3 .And, how RELEVANT or IRRELEVANT is the COMPETENT/INCOMPETENT distinction to your conception of BIRCH BAYH? IRRELEVANT: : : : : : : :RELEVANT 3 2 l 0 l 2 3 (BIRCH BAYH) EXPERIENCED: : : : : : : :INEXPERIENCED 3 0) M p 0 rd '0 [C K/ 67 68‘ 69 70 71 72 73 7Q Now, how IMPORTANT or UNIMPORTANT is the EXPERIENCED/INEXPERIENCED distinction to your conception of the IDEAL CREDIBLE SOURCE? UNIMPORTANT: : : : : : : :IMPORTANT And, how RELEVANT or IRRELEVANT is the EXPERIENCED/INEXPERIENCED distinction to your conception of the IDEAL CREDIBLE SOURCE? RELEVANT: : : : : : : :IRRELEVANT ATTRACTIVE: : : : : :> : :REPULSIVE Now, how IMPORTANT or UNIMPORTANT is the ATTRACTIVE/REPULSIVE distinction to your conception of the IDEAL CREDIBLE SOURCE? ’ IMPORTANT: : : : : : : :UNIMPORTANT 3 2 l 0 l 2 3 And, how RELEVANT or IRRELEVANT is the ATTRACTIVE/REPULSIVE distinction to your conception of the IDEAL CREDIBLE SOURCE? IRRELEVANT: . : : : : : : : RELEVANT INTELLIGENT: : : : : : : :UNINTELLIGENT Now how IMPORTANT or UNIMPORTANT is the INTELLIGENT/UNINTELLIGENT distinction to your conception of the IDEAL CREDIBLE SOURCE? IMPORTANT: : : : : : :‘ :UNIMPORTANT And, how RELEVANT or IRRELEVANT is the INTELLIGENT/UNINTELLIGENT distinction to your conception of the IDEAL CREDIBLE SOURCE? RELEVANT: : : : : : : :IRRELEVANT -1u- ill _\.I I, 75 76 77 78 79 80 (IDEAL CREDIBLE SOURCE) JUST: : : :v : : : :UNJUST Now,tunvIMPORTANT or UNIMPORTANT is the JUST/UNJUST distinction to your conception of the IDEAL CREDIBLE SOURCE? IMPORTANT: : : : : : : :UNIMPORTANT 3 2 l O l 2 3 And, how RELEVANT or IRRELEVANT is the JUST/UNJUST distinction to your conception of the IDEAL CREDIBLE SOURCE? IRRELEVANT: : :’ : w : : : :RELEVANT RELIABLE: . : : : : : : :UNRELIABLE Now; howIMPORTANT or UNIMPORTANT is the RELIABLE/UNRELIABLE distinction to your conception of the IDEAL CREDIBLE SOURCE? IMPORTANT: : : : : : : :UNIMPORTANT And, how RELEVANT or IRRELEVANT is the RELIABLE/UNRELIABLE distinction to your conception of the IDEAL CREDIBLE SOURCE? IRRELEVANT: : : : : : : :RELEVANT 3 2 l O 1 2 3 -15.. DL Dup 1-6 Card 11 7-8 9-11 13-15 17-19 21-23 25-27 29-31 y If 0 represents total lack of competence and 100 represents the level of competence of the most competent person you know personally, HON COMPETENT IS BIRCH BAYH? (Use any number to report as accurately as possible how competent you think Birch Bayh is.) If 0 represents total lack of experience and 100 represents the level of experience of the most _experienced person you uknow personally, HOW EXPERIENCED IS BIRCH BAYH? . If ()represents total lack g: attractiveness and 100 represents the level of attractiveness of the most attractive person ygu know per- sonally, How ATTRACTIVE IS BIRCH BAYH? If 0 represents total lack of intelligence and 100 represents the level of intelligence of the most intelligent person you know personally, HON INTELLIGENT IS BIRCH BAYH? If 0 represents total lack of justness and 100 repsresents the level of justness of the most just person you know personally, MON JUST IS BIRCH BAYH? If 0 represents total lack of reliability and 100 represents the level of reliability of the most reliable person you know personally, HOW RELIABLE IS BIRCH BAYH? -15- Finally, we would like some information about you. There are several questions below that we would like you to answer. If, for any reason, you do not want to answer one of more of these questions, please feel free to skip it. How- ever, try to answer as many questions as possible. Let me remind you that this information is kept strictly confidential. It is needed only to compare samples of subjects who are filling out this questionnaire and other question— naires similar to it. Again, thank you for your time. 1. What is your age? 2. What is you sex? Male Female . (Circle one) 3. What is your year in school? Freshman Sophomore Junior (Circle one) Senior Graduate student B. What is your major? (If no major, please write "none") 5. Race. American Indian Black Caucasan Chicano . (Circle one) Oriental Other (please specify) -17- . , .f.\_ \ ll. «w