o . l o . ... .. . . 9‘ .- o . .o . .. . h . - . . . .. . . . . .o . . .. . . . . a ‘ fl! IUI m x EN’RLI A CAS.: 8" A-LYSES 0F TS . MEN UULCM * K TOR AN DIFFERENTIAL JUDG Utmmn r 0? FAQ "J“ U!“ U55 SEM‘ANTIC U! . . ‘ ‘I . . v .1.” n., .... .3 .. ....GL; 0.. .v Thesis for the Degree of M. A. UNWERSITY . ‘J. MICHIGAN STATE V . . 0.I..—.y.404§4Vo‘OOAO.O¢¢QC‘Qo .. V . .04... o . 0 7'0 . o , .. rt..........l.o..o.-o oroco on. ”/3.” . . . I. .. ...,ovn... ... . . 9... so. . .0”t.f ... . . ..... . c.......,.o . 9,-0.1. . . . . .a V o .....uoa ..tfv . o .I. . r . V _ . . V. .. ....V ..41. ... V c . ¢.'.'.....p.o v; . . . . —’IO> .. .......:_... 22.. a D . . . 4... n _...'..vot.‘u.c . . .o. ....Ol: . 0.. o _(oo 1.. I.... . . . o!.oVVO.~ 1.. If: . .. . /o.'..-I. ’ V _ . ..b.o .0/ 2.1... _ . ¢. . t..c-a—. .-O_l.Iv-Ioooo. _ . o v . ‘ p .u . . . .....x: 43.... v: o .. r..;....a .. . A . . .‘ o ...;..4... don't/...... .o /. . .....IvI .. o . . .l. ., ...... vat.’ . ro_ou .Ino. Q.‘¢.c0..0-O, . . . . , ..4 o ....1. a Dr 90.... . . . .. I. . . .. - .....ro._H......-.d...’.. ....1'..o'. . .. . V I _ . .. ... . .V .. .. .a u . . . til 4... . V. . .Oo. 1.... . c. 0!. W _ . ,.. ‘ ... .r . a a . n _ .. ..,.¢c. ,.... .’. «24.....0. .a.. _ . . . . . ..9. I... ‘u I ..v... 1.-...4 . J ... u 'v. ’u..'. . p.....' ......naro {...-vff y . . .. ... ....oo.;lto . . . _ . .. . . . .J . . 4 n, ... . . . . , . . . 3,. . .. x. . 1a 1.3.... {...-.....rw... ..- ....l .. . .. - ... . .1... . .o to. o....c..lo' o.lo 'ufl’: 0. l . . o . : 0" .....v ._ 9¢.o...o- ... _ .c n u . a ..v. .. . .. . ...........J. .ooc./ .1 .fp.o.o.r1r. V A . . . . ..r. ..70; .90. 2”...“1 . . . .. n . . .rc.' 1!.v...... o. ....u o . ... . . D _ . . . .. _. .r; o ...o. ..,.ai....<’.¢. ....v.’ .1 . . . . . . I ., .o . r v. . .1 4. .- .VJJ.. alou...‘. Oflbnply: 9’ . cc . .‘A'.o “:10 a o y. u. . . . . .. u . 3'90 .../l o......'.'oo(d. .. J . . . a. . .1 IV. ...-...p: (.1. '1 0V . . . . . . . . .o 0 ¢ a. . . 0.. . b ...,n. .. .m . . .. . ... . . , . . o..._........o:o. .tp..o|./l.v-..a.. 'O/ \ . d , .o. . . '. ... n .. . V o . o . . . . _ . V . . .. . . V. .. w I u t l . .. ... ... . V .. .19.. A . .V o. . . . . .. . _ f..o.o... ' V . a _ - . . . .. V . ... .n ... J (I .904 o d _ - _. . . ...... . . up..,.V...a m . . . .. . , 2. a: L . _ . . .o . . .. ..J. o $.11 . .. ... .. ... ...A. .— . . . . r4; .... 4 ... o I.VI. V . ... . O . . .tt ,- a . . :4. . .o n . II a .O h f '0 '.'n 'p . .. . V . l ./..p .1'c.r V . .. . V ..;,,.V. . ‘ V. . . . V .. . . .. ..... , _ . .l. J . . ... .,. V... n. . o . a O u .. .Hp. . rd . ._. v.9 ‘ a.. . _ . a... . . . .. . . «r. . a . . . . r u .. J... . _ 0.. _ n 11. .’ . . . . I alu.al . 196:8 o. . . x . r . o . o. .. . 0...‘. v . . . J . on. t. v ; . 0 o v . J a . . . .. V. . . I . . — V . a y 4 ‘ . .I v: .I I . .. O l ..t a I. 9 o _ . u _ V ..I o a V. u. ain .1 . . . . r J 7. at . . a - d a - . ... . n 1.4.... . . . o. . I I o. C. o v d . . . . o .f.. . . V . .r . ... . _ o . n ,u . I I o o o o o n (0 ~ . a o - ... v .. v o4 .c O y . . l. .. n. It. ...! o . - . _. _ I. .Jo . .. . n . I . o . . . l . ; .tr -. . v t . ...: I n . . . . or . . V . . 4' . .a n . .- . . .. u .. V o r . y . ... .9 .. . . r _ a . . .- v . no . I... . v. I v. o .A .. . . o .. I. . . .u r c 0.00 . t l-.. c. . . o.“ r . . . . . n o O l a l . . . I . . . .t. J o O u l o v o . o o o o p . a V < O - . ‘ . o . u U .’ n . l l t - ' C .u o l A O ' n I . . . . p| . . s .. . . .v . . . 0’ . o r u 4 O u - ' .h _ I n o I v C . A . /:r.o .. . . . . p. u . . . - . . n f . r a. o . fly. . IV. 1. O O ...... .n... n 1. ... . c. .. .' s x.~:.did .U§ ‘AI 0‘ ..§ 0 I . o; I. I .. 1 no. I O . . I h n t” . .s . ... . . a. v. .. N.h...........1t.. ...? .....J o o. 33.. -. .Vauybs :“u..tnu “$.Iu “AWE?” vb .o ”#“r 1.11. r9 9.. .. I.........Q..u...o..fi_..—. ...-... .J_.oIVA¢....c . .....Oo 3... §.m¢.b.n~..oo‘«.~o.bo‘ 0... ; .4 (0."1‘ 0;. Col x... I ‘ -.. "i I! ’ III 1|.- it \\ 3 1293 009 7 m '\ \ WWW L {Hti'tl‘i ’W W -“M MSU RETURNING MATERIAL§: Place in book drop to LIBRARIES remove this checkout from am your record. FINES will be charged if book is returned after the date stamped below. I“; 2': 5 “a 5'11: 1) “J i _, _L w 76‘] ABSTRACT DISCUSSION OF CERTAIN PROBLEMS IN INTERPRETATION OF FACTOR ANALYSES OF SEMANTIC DIFFERENTIAL JUDGMENTS: A CASE STUDY by John W. Powell, Jr. Factor analyses of semantic differential (SD) judgments present many problems of interpretation and generalization to other concepts and respondent groups. There appear to be many processes at work which act to reduce generalizability of factor structures. Differences or contradictions from one study to another may be due to 'real' differences in the populations of concepts and respondents studied, or to procedural and interpretational differences, or to uncontrolled 'irrelevant' population differences. These problems include concept/ scale interactions, information losses due to summing over many concepts and reSpondents, and to the vaguaries of the factor comparison process. In the present study, six respondent groups (N=31/group) each judged five different family planning concepts on 90 SD bi-polar adjective scales. All groups used the same 90 scales. The concepts represented three a_priori domains within the general topic area of family planning: 1. contraceptive techniques and devices 2. information sources, and 3. certain issues concerning birth control. Each group judged at least one of each type of concept. Demographic information was also collected from each respondent. Factor analyses of the adjective scales were produced from each group in three waysi Mode 1, in which responses to all five concepts in each group were summed; Mode 2, in which only reSponses to 'technique' concepts were included in the analyses; and Mode 3, in which only responses to 'source' concepts were analyzed. Differences in the results of these analyses might occur (a) between types of concepts (modes) within a group (b) between groups for the same type of concept. If structural differences occur between types of concept, this would be evidence of the specificity of semantic dimensions to various classes of concepts. If across-group structural differences occur, then group demographic or personality variables may be responsible, or the concepts though to be 'similar' were perceived as different by respondents. Thus the design of the present study allows observation of the SD 'in action' and conclusions may be drawn concerning the range of variables which affect factor structures and generalizability of the dimensions of any single SD study. The method of comparing factors from various solutions or studies can greatly affect conclusions based on the studies. In what might be called the 'semantic' method, the experimenter attempts to interpret the nature of a factor from examination of the highest- loaded scales. Two factors are seen as related if the top scales on one factor are the same as, or are synonymous with, scales on the other factor. The 'correlational' approach uses some statistical technique, such as an index of factorial similarity, to rate factors as same/not same. In the present study, both methods are utilized and the conclusions resulting from each are compared. Results indicate considerable variation in factor structures both across groups and across analytic modes. Observation of highly loaded scales suggests that a three-factor solution, whose dimensions were characterized as 'Social,‘ 'Functional,‘ and 'Personal' Evaluation, describes responses in all groups, in Modes 1 and 2 (mixed concepts and technique concepts). This structure is probably most applicable to 'technique' concepts. The Mode 3 (source concepts) solutions exhibited such a high degree of variation in scale usage that no overall pattern was discernable. This implies that concepts here categorized as 'sources' are psychologically quite heterogeneous. The correlational factor-comparison method (PAC-SIM) indicated that a one-factor 'Evaluation' solution with many idiosyncratic factors was the most conservative description of the results. It was decided that without elaborate re-analysis of the data, no definite statements could bermnkrlinking demographic differences among groups to factor structure differences. 4 //77 )3 DISCUSSION OF CERTAIN PROBLEMS IN INTERPRETATION OF FACTOR ANALYSES OF SEMANTIC DIFFERENTIAL JUDGMENTS: by John W. Powell, Jr. A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Communication 1968 A CASE STUDY Accepted by the faculty of the Department of Communication, College of Communication Arts, Michigan State University, in partial fulfillment of the requirements for the Master of Arts degree. t3 ‘ 7 (,5? _. 1,. [ _/}rt{¢1--\, Qf- ilklft.u/ “4*”; Director of Thesis ACKNOWLEDGMENTS I would like here to record my thanks to the following peOple who made this thesis possible: Dr. Erwin Bettinghaus ; Thesis director, and my all-purpose advisor. Dr. Larry Sarbaugh and Dr. Randall Harrison; thesis committee members and patient friends. Barbara Judy, who worked many hours data collecting, coding, and morale-building. Anita Immele and Betty Darlington, who somehow managed to follow my incoherent instructions about computer analysis. Bill Tedrick, who was responsible for making most of our 83 available to us. Bridget Roberts, Barbara Bernath, and Claudia Gean, who, through weddings and weekends, converted my very rough manuscript into a thesis. TABLE OF CONTENTS CHAPTER I INTRODUCTION AND RATIONALE . . . . . Introduction . . . .-. ... . . . Purposes of the Present Study . . II PROCEDURE . . . . . . . . . . . . . Selection of Concepts . . . . . . Selection of Scales . . . . . . . Relevancy of Scale Adjectives to Co Test of Adjective Antonymity . . . Questionnaire Construction . . . . Pretest of Questionnaire . . . . . The Respondents . . .-. . . . . . Analysis of Data . . . . . . . . . III RESULTS . . . . . . . . . . . . . . Variance Accounted For By The Solut Factor Identification & Interpretat Factor Relations Indicated by FAC-S Technique . . . . . . . . . . . Effects of 'Evaluative Opposition' Group Demographic Differences . . Similarity of Factor Structures to and Berlo Models . . . . . . . Item Reliability . . . . . . . Summary of Results . . . . . . . . IV CONCLUSIONS . . . . . . . . . . . . Discussion . . . . . . . . . . . . Suggestions For Future Research . REFERENCES 0 o 0 O O O O o o o o o o o o o 0 ions . ion . . IM Osgood 14 14 15 16 17 17 19 20 21 3O 3O 34 39 41 43 44 46 47 49 49 56 59 TABLE LIST OF TABLES Percent Variance Accounted For by Factors in 2 ,. 3 And 14 - Factor Solutions; Mode 1 (Mixed Concepts) Analysiloooooooooooooeooooooooooo Common-Factor Variance; 3 - Factor Solutions; Mode l . . Percent Variance Aecexmted For by Factors in 2, 3, Andh-Factor Solutions; Mode328c3 . . . . . . . . . o Common-Factor Variance; 3 - Factor Solutions; HOstz&3ooooooooooooooooooooooo Scales Common to Three or More 'Social,’ 'Fanctional,‘ And 'Peroonal' Evaluation Factors Within Mode 1 And MWZOOOOOO.....OOOOOOOOOOOO... Scales Common to 'Social,‘ 'Functional,‘ And 'Peroonal' EvaluationFactorsAoross Modes 1&2 . . . o . . . . . . ”proximate Percent of Variance Ebrplained by First Five (of 90) Dimensions of Principle-Ins Solutions; Mode 1 MSOBOoooooooooooooooooooooooo Factor Sinilarity Coefficients Comparing 3 - Factor Solutions for the Concepts Your Fang Musician hdSW1‘1WG1‘kGr8oooooooooooooooooooo 30 31 32 33 36 38 no Figure Figure Figure Figure 4. Figure Figure Figure Figure Figure LIST OF FIGURES Figure Showing Possible Comparisons Among Factor Analyses . . . . . . . . . . . . . . . . Concepts Appearing in Each Questionnaire Form . Sex Distribution of Reapondents, Across Groups. Marital Status of Reapondents, Across Groups. . Average Number of Children Per Person, Across Groups. 0 O O O O O O C O O O O 0 O O O O O O O O Reapondents' Home Residence Area, Across Groups Education Level of Reapondents, Across Groups . Age Distribution of ReSpondents, Across Groups. Factor Analyses Performed in Each Group . . . . Page 11 18 22 23 24 25 26 27 29 LIST OF APPENDICES Page APPENDIX A Questionnaire Instructions Sample page of questionnaire . . . . . . . . . . . . 61 APPENDIX B Demographic Questions Listing of Demographic Characteristics of Reapondents, By Group . . . . . . . . . . . . 63 APPENDIX C List of Adjective Scales Appearing in Questionnaire . . . . . . . . . . . . . . . . . . 67 APPENDIX D Scales Selected by Factor Loading to Represent Each Factor - 3-Factor Solutions - Modes 1, 2, & 3 ... . . . . . . . . . 69 APPENDIX E Factoria1 Similarity Coefficient (FAG-SIM) Tables 0 C . O 0 O O O O O O O O C C O O O 0 O O 87 APPENDIX F Adjective Emitted to Six Family Planning Concepts . 96 CHAPTER 1 INTRODUCTION AND RATIONALE Since its introduction in 1957 by Osgood, Suci, and Tannenbaum,l the semantic differential has stirred broad interest and has been applied with varying degrees of success in many and diverse fields. Essentially a paper- and-pencil rating task, it is easy to administer and yields masses of data very quickly. Its difference from other rating scale techniques, and its raison d'etre, lies in its relation to Osgood's theory of affective meaning as implicit mediated responses to stimuli. I need not here attempt to explain or criticize that theory; it has received considerable attention elsewhere.2 Suffice it to say that the semantic differential (SD) technique is held to be a means of making eXplicit and observable some aspects of the implicit mediation process. In brief, the technique employs seven-point linear scales bounded at the ends by bi-polar, or Opposite, adjectives. These scales are then used by $8 to rate various concepts, objects, or events. 83 use check-marks to indicate which adjective of each scale best describes the concept and 'how much' it does so - marking thus varies in 'direction' and 'intensity.‘ By applying factor analysis to the intercorrelations among scale responses over many 83 and concepts, independent clusters of scales (factors) which represent the same kind of response may be located. These factors are then said to indicate the major dimensions or axes in a 'semantic space.‘ There could be any number of such dimensions. Osgood had, apparently, no hypotheses about how many semantic dimensions there might be, but based on the results of his and others studies, he has settled on the assumption that there are three major dimensions. These -2- three dimensions he has characterized as 'Evaluation,’ 'Potency,' and 'Activity,' and they usually explain from 50% to 60% of the variance in scale responses. Having identified recurrent semantic dimensions, Osgood is then able to select a few scales from each which will represent them and to construct a brief instrument which should be capable of recording most of the relevant affective responses of any S judging any concept. This technique, at one stroke, should allow statements about the structure of affective response and provide a handy instrument for discovering differences in 'meaning' among diverse objects or for separating Ss in terms of the 'meanings' they have for the same objects. As always, things are not as straightforward as they ought to be: There are many important problems to be faced when using the SD as a measurement technique. Osgood has been at pains to point out the number of assumptions underlying the SD which may not, in practice, be met, and the problems encountered in reaching hard and fast conclusions about factor structures and identity of scales with factors. However, many of those who have employed SD in their studies have chosen to take these assumptions as met and the results of their own experiments, then, as generalizable. Compared to the studies which adopt the SD willy-nilly, there have been relatively few studies whose aim was to validate the assumptions or explicate the basic prdblems contained in the SD technique. This study, while not solely aimed, or solely designed to test aspects of SD methodology, does offer an interesting Opportunity to examine the SD in action. Viewing the SD in the context of factor analysis, there appears to be a complex of processes which may combine or interact in obscure ways and which act to muddle the clarity of solutions. These sources of -3- variation both alter the probability that a particular analysis will be interpreted 'correctly,' and the probability that similar outcomes will be recognized across different analyses. Differences or contradictions in findings from one study to another may be due to 'real' differences in the populations studied, or to procedural and interpretational differences, or to uncontrolled 'irrelevant' differences in the pOpulation, or to any combination of these. If the semantic differential is going to continue to be widely utilized and to contribute to our knowledge of human psychologi- cal processes, then it is important to separate the wheat from the chaff, the real differences of concern to theory from the noise of unaccounted- for variance in our results. A manifestation of the uncertainties imposed by methodological diffi- culties in SD research is the Relativistic y§_Absolutistic argument concern- ing the nature of semantic space.3 Many factor analytic SD studies have concluded that there are three major dimensions of affective response which may be routinely identified as Evaluation, Potency, and Activity. These factors have turned up even across different language/culture communities.” The 'absolutistic' conclusion based on these findings is that peOple demonstrate a constant meaning structure despite differences in placement of referents within the structure. On the other hand, there are studies--usually dealing with a restricted range of concepts--and in some other language/culture communities, in which factors emerging from SD judgment are not recognizably the same or are severely modified versions of the 'standard' dimensions.5 These results imply the relativistic hypothesis that humans will apply various standards of affective meaning depending on the stimulus and their own backgrounds. -u- If one problem area should be singled out as the most troublesome, it would probably be that involving concept-scale interactions. The factor analysis of SD judgments is intended to allow generalizations about semantic space which pertain to the universe of potential concepts. Unfortunately, there are many instances in which scales load highly on a factor for concept A but do not do so for concept B; the scales may either be irrelevant to B or appear on some other factor altogether. That is, scale intercorrelations show variation across concepts. Since the dimensions of semantic space are identified by the scales which load on them, this sort of scale instability causes considerable confusion. Furthermore, the polarity of scales may change from concept to concept. The adjectival scale end which is favored (positive) when judging concept A may not be the same scale end favored for some other concept, although the scale's factor identity is unchanged. This phenomenon contributes to confusion as it implies that the polarities of the dimensions of semantic space are not fixed but vary with the concept. Another aspect of this same prOblem is what might be called 'evaluative opposition' of concepts on certain scales. Whenever two or more concepts are thrown together in analysis to yield a single common set of scale intercor- relations, a potential source of error lies in the possibility that judgments for concept A may diametrically oppose judgments for concept B on many scales. This could result from a polarity reversal or from Opposing evaluations of the two concepts. If such Opposition occurs with any regularity in the data, then the averaged judgments on these scales will tend to 'cancel out' and lead to artificially low correlations and obscured factor structure. Scales which might make important discriminations about either concept alone may not appear at all when judgments on concepts are summed.6 Osgood has posited two mechanisms which he feels account for most of the instability in scale behavior; these are termed 'denotative contamination' and 'factorial coalescence.‘7 The former occurs when scale adjectives are interpreted differently by Ss in conjunction with different concepts. Words can have many meanings and there is no way to control how 83 will 'use' the adjectives. These differences may be between literal and figurative (donotative and connotative) use, or between various figurative uses. The practical result is that 83 use the adjectives as homonyms, (words Spelled the same way but having different meanings) while in interpreting the result- ing factors we must assume that scale adjectives have only a single meaning. Factorial coalescence makes use of OSgOOd's Congruity Theory notions to describe the interaction of a scale with a concept. Both elements of this dyad possess connotative meanings and although the scale is intended to index the meaning of the concept, the resulting judgment is a compromise position between the meaning of the scale and that of the concept. In cases where the concept has strong evaluative or dynamic properties associated with it, the judgment compromise will tend to shift scale markings toward the dominant concept dimension, again, obscuring lines of demarcation between factors, e.g. the merging of activity and potency factors into a single dynamism factor in some studies. Summing or combining judgments of different 83 on different concepts presents other difficulties. Talbot has written: Summing across both Ss and concepts implies using all concept ratings on a particular scale for each 5 as indi- vidual observations in computing the correlations. Then the number of observations would be the number of Ss times the number of concepts for the purposes of correlation computation. These same procedures were employed in the work Of both Osgood and Berlo. The factor structure under such conditions could have been derived from variance attributable to Ss, concepts, or an interaction of both.3 And, according to Darnell: In the normal SD analysis, there are two kinds of variance contributing to the outcome. Variancel is the dispersion of scores around the concept means. Variancez is the dis- persion of concept means around the grand mean. Variancel and variance2 are independent in that neither is predict- able from the other on §_priori grounds. Thus there is no reason to expect that a factor analysis based on either kind of variance alone would be the same as one based on both . . . . These findings suggest strongly that the factor structure is heavily dependent on V2--the among- concept variance.9 Generally, researchers are concerned with responses to individual con— cepts although their method of dimensioning semantic space lumps concepts together. The alternative extreme procedure would be to do factor analyses for each concept separately to avoid the kinds of 'noise' described above. This approach would, however, drastically reduce generalizability of results and multiply the amount of effort required to sample an area of meaning. It reduces SD technique to a low-efficiency descriptive method. A compromise alternative is to do separate analyses for concepts from one coherent topical area, concepts which are all 'alike' in some demonstrat- able way. Osgood has said: What do these findings have to say about the practical problems of semantic measurement? For one thing, it now seems less likely that we will be able to discover a single set of scales which represent an adequate set of factors and which are stable across whatever concepts may be judged. On the other hand, it may be possible to identify classes of concepts for which general instruments may be used, and perhaps, in course, the principles which operate in determining a common frame of reference can be dis- covered.10 In the context of the present study, the implication is that although 'Evaluation,‘ 'Potency,' and 'Activity' scales of Osgood's studies may not be relevant to family planning concepts, perhaps we can locate other sets -7- of scales which ARE relevant to these concepts. Or, more likely, we may find sets of scales relevant only to subsets of family planning concepts, such as birth control devices or information sources. This approach also suffers difficulties. The basis for deciding which concepts are 'alike' is not easily perceived. Concepts which are 'alike' in that they are drawn from a common area of experience or represent the same class of objects may still not be 'alike' psychologically. The classes of concepts of interest to researchers may not correSpond to psychological concept classes . Another area of concern and source of confusion, especially when com- paring the results of different factor studies, is the role that 'people differences' play in determining meaning structures and/or factor analytic results (they may not be synonymous). Two realms of people differences are, roughly, demographic variables, and personality variables; these are probably overlapping categories. We might expect that different life experiences, as indexed by demo- graphic information, would lead tO natural variation in development of the possible qualities of an individual's semantic Space, e.g. urban ghetto dwellers might demonstrate consistently different semantic dimensions than would WASP upper-middle class populations, and so on. In terms of the present study, we might expect that relevant aspects of contraceptive techniques would differ for inexperienced, unmarried undergraduate women and for middle-aged housewives. For the present study, demographic infor- mation was collected from each reSpondent covering such areas as age, family income, education, religion, marital status, number of children, ideal number of children, residence history, and preferred residence. If there are 'large' differences between respondent groups along some of these variables, it may be possible to suggest relations between them and vari- ations in the groups' structuring of the SD scales into factors. Another 'people' difference of concern are personality traits of the re8pondents which, in a small sample, may be non-randomly distributed and capable of altering the outcome of the analysis. Talbot has indicated that there are likely to be structural differences resulting from the S's per- ception of, in his study, a 'source' as having a position similar to or) different from the S's own position. Talbot has alSo underlined the need to discover the personality correlates of differences in semantic dimensions.ll We did not collect any personality measures in the present study. A concern on a different level is how best one may compare, or estimate the similarity of factors occuring both within a solution and across solu- tions and studies. Within a single solution we need to ask if the factors obtained are different enough to be considered as separate and truly independ- ent factors, or if they are merely arbitrary subdivisions of some larger di— mension. Perhaps more importantly, how best can we determine that two or more factors obtained from different groups or even different studies represent the same theoretical factor? This question is a crucial one in that the ability of semantic differential factor analytic research to provide substantiation for theory requires that there be a way to establish that results have been replicated or not replicated in other contexts and with other 33. The practical use of SD technique also requires that results of various studies be comparable. Two possible methods of factor comparison might be termed 'semantic identification' and 'correlational identification.‘ These two approaches do essentially the same thing but go by different routes which Often lead to different conclusions. 4 The 'semantic identification' route emphasizes that factors must make sense, that is, the items occurring in the factor must clearly and consistent- ly represent a meaningful kind of response to concepts. Given that we can (intuitively) identify the nature of a factor, or the kind of reSponse quality it represents, then other factors are similar if they seem to make the same kinds of distinctions about concepts; e.g. evaluation, potency and activity dimensions are often identified even when many of the top-loaded items are not ones appearing in other studies. Ideally, of course, the same scales have been used and appear grouped in the same ways as in other studies, but recurrence Of the identical scales need not be a requirement so long as their 'intention' is the same. This kind of decision is personal and judg- mental; correctness could only be indexed by the amount of agreement among judges. That a factor does or does not Jmake sense' (is recognizable as some reasonable and consistent question about concepts) is also used as a criterion of whether a factor is the result of systematic or random processes occurring in the data. An alternative approach which attempts to do away with intuitive or personal judgment of factor identity is the correlational technique known as the coefficient of factorial similarity, whose formula is as follows: I fl f2 Q f15%22 where fl f2 is the sum of the cross-products of the factor loadings of any two factors and fl2 and f22 are the sums of the squared factor loadings.12 This technique provides a measure varying between -l.00 and +1.00 like a Pearson Product Moment Correlation Coefficient. It requires that the same -10- scales (variables) be employed in all factors to be compared. For this reason, it has found wider application comparing factors within solutions of the same study as an aid to determining when to terminate rotation than in comparing factors across studies, where the scales and concepts are likely to be different. Since there is no sampling distribution available for this statistic, significance and confidence intervals cannot, in the usual sense, be computed.* In practice, then, the factor similarity coefficients still provide indications which must be interpreted by people. The kind of information it yields, the 'similarity' it describes, makes use of all the variables listed on a factor. This is logical but conflicts with the more traditional practice of characterizing a factor only by the few scales with the highest loading or purity and ignoring the rest. One result of this difference is that factor characteristics which appear important and convincing from inspec- tion of top-loaded items are washed out in FACSIM comparison. Purposes of this study: Three main problem areas discussed were the range of concept/scale interaction phenomena, methods of factor identifi- cation and comparison, and individual 'people' differences. The present study offers an Opportunity to examine the effects of the problems discussed above and their influence on structures resulting from factor analysis of semantic differential data. Concept/scale interactions: We expected that the six factor structures occurring within each analytical mode would resemble each other.** This *A 'rule of thumb' guideline exists termed the lower limit of best fit: 1 + l/k 1/2 where k is the number of factors in the solution. For three- 2’ factor solutions this value is 0.788. ~ Six groups of semantic differential data were collected. (n=32/group) -11- expectation constitutes our practical null hypothesis. In general, we also might expect that Mode 1 structures (mixed concepts) would tend to resemble 'Osgoodian' space with evaluation, potency, and activity dimensions--as much as allowed by the scales we employed. Mode 3 (sources) structures we might expect to reflect dimensions found by Berlo et al for evaluation Of message sources--safety, qualification and dynamism.ll For Mode 2 (tech- niques) we had no prior model. Dimensions postulated a_priori as reasonable for these concepts included 'ease,' 'effectiveness,‘ 'understandability,' or 'personal evaluation.‘ Group 1 2 3 4 5 6 Mode 1 (Mixed) A .---4---0---.---.-) Mode 2 (Techniques) Only U‘ <-—----: Mode 3 (Sources) Only If differences occur among structures they may occur (a) between groups in the same mode of analysis (b) between modes in the same group (c) or any combination of a and b. (See diagram above.) We assumed that differences would All groups used the same 90 scales but rated five different concepts each. The concepts represented three domains within the general area of family planning--contraceptive techniques, information sources, and 'issue' concepts. Thus, there were levels of expected or face similarity among concepts. Each group rated at least one of each type of concept. Factor analyses were produced from each group in three ways: Mode l--with all five concepts summed over all 85; Mode 2--in which 'technique' concepts were pulled out and analyzed separately. If two techniques appeared in a questionnaire they were summed; Mode 3--in which 'information sources' were pulled out and analyzed separately. If two sources appeared in the same questionnaire, they were summed. -12- be greater across modes than across groups within a mode. Either 'vertical' or 'horizontal' structure differences could indicate concept/scale inter- action. 'Vertical' changes (between modes) would clearly demonstrate the specificity of semantic dimensions to classes of concepts. These changes, if they occur, should reflect meaningful or reasonable alterations in structure to match the nature of the concepts being judged. 0n the other hand, we might expect that differences occurring 'horizontally,‘ or across groups within a mode, would result from more random kinds of concept/scale effects since five different concepts appear in each group, and from the uncontrolled differences in group demography and personality, producing subject/scale interactions. Thus, the changes which might occur would be less likely to 'make sense' except as functions of unique circumstances of subject/scale and subject/concept interactions. Factor identification and comparison: The general plan of attack will be to work withfactor loadings and purities initially and attempt to map in the identities and relationships of factors as well as possible in this way. Then, with these 'semantic' conclusions as predictions, FACSIM coefficients will be examined. The pattern of factor relationships which this correlational method indicates will be compared with the prior con- clusions. The question of interest is what major differences in interpreta- tion does each method introduce and when is each appropriate? People differences: We collected what we considered to be pertinent demographic information from each respondent, as described above. Although some effort was made during data collection to keep the groups' makeup parallel, demographic variables were not closely controlled and inter-group differences were bound to appear. By comparing demographic patterns of the AAA-(i -13... six groups with variations in factor structures, it may be possible to make some statements about the relevance of such variables to SD factor analytic studies. Because in this study, the effects of any demographic differences could be confounded with effects due to the unique set of concepts judged by each group, conclusions in this realm must remain speculative. To reiterate, the questions to be examined in this study which relate to the methodological prOblem areas discussed are: (a) how common were factor structures between groups and between §_priori concept classes (modes)? (b) was there agreement or disagreement between 'semantic' and correlational factor comparison techniques? (c) did the demographic dis- parities between respondent groups appear to influence the resulting factor structures? 14 CHAPTER II PROCEDURE In this chapter, the procedures employed to generate the measuring instrument, gather the responses of subjects, and analyze the resulting semantic differential judgments, will be described. The topics to be considered are (a) selection of the family planning 'concepts' to appear in the study (b) selection of the adjectival scales which all respondents use to judge the concepts (c) the allocation of concepts to the six 'forms' of the questionnaire (d) the construction of questionnaires and the format employed (e) results of a pretest of the questionnaire (f) the selection of respondents in the six groups and demographic comparisons between groups (g) the analyses performed on the semantic differential data. Selection of Concepts: Thirty concepts were employed representing three areas within the family planning sphere--(l) general evaluative "issues," (2) specific contraceptive techniques, and (3) sources of information about family planning. They were the following: Area l--"Issues" birth control intentionally limiting the size of your family large families the population explosion the teen-age parent religious rules against birth control measures state laws against distribution of birth control devices U. S. birth control projects in other countries pregnancy before marriage Area 2--Techniques the birth control pill the intrauterine device (IUD) the rhythm method ‘ | ‘1 L' '1 ll Iv l! l [l all! I‘ll I“! ll I'll. lililll‘l III. {I‘ -15- Area 2--Techniques the cervical cap sterilization of women sterilization of men legalized abortion the diaphragm a birth control pill for men sterile foams, gels, or creams the condom Area 3--Information sources sex education in schools the Planned Parenthood Association your family physician good friends of yours articles on birth control in newspapers and magazines birth control clinics the Catholic church social workers your parents your minister or religious advisor Osgood has said, "Ordinarily in making up a sample of concepts for a differential we try to balance off good concepts with bad, strong with weak, and so forth. . ." (1, pg. 85). Other criteria are: applicability of the concepts to the problem under study and the adequacy of sampling of relevant concepts. The concepts employed in this study vary in a number of ways: in complexity, assumed favorability of evaluation, 'concreteness,' familiarity to respondents, impersonality, and so on. So although all the concepts are confined to the Special context of family planning, they are non-homogeneous in many respects. Scale Selection: Our computer facility handles up to 90 variables in factor analysis. Because we were 'mapping' a new semantic area, it was decided to include in the study as many scales as possible, so our target was 90 acceptable adjective scales. -16- Rather than borrowing verbatim all the scales used by Osgood in the Thesaurus study, or by Berlo et al. (5A) in the Source Evaluation studies, an original list Of adjectives was compiled. Roget's Thesaurus was the initial source for terms and apprOpriate antonyms. The major selection criterion was relevance to the concepts already selected. The majority of adjectives chosen represent 'evaluative' aspects of judgment, rather than 'dynamism' aspects.* After the final list of scales was chosen, it was felt desirable to obtain an independent check on the relevancy of the scales to potential family planning meaning dimensions. To this end, 33 male and female undergraduates were asked to supply adjective associations to three ’technique' concepts and three 'source' concepts. The adjectives were presented in pairs, each S responding to two concepts only. 0f the 342 adjectives (260 different terms) supplied in this fashion, nearly all were 'evaluative' in the same way as the adjectives already selected. Only seven adjectives suggested size or activity in ways not occurring in the study; they were: growing, hard, huge, minute, small, dainty, aggressive. These seven terms were emitted by five 55 judging two of the six concepts. In many cases, adjectives were clearly denotative descriptions of concepts rather than connotative reactions, e.g. THE PILL: large-small; MY MINISTER: nonexistent.** But, on the whole, the adjectives employed in the study closely resemble the terms which were associated to *The final 90 scales include many used by Osgood and by Berlo. These are indicated on the list of scales in the appendix. **This tendency to respond denotatively may have been produced by the instructions accompanying the questionnaire which asked 83 to list adjectives which 'could be used to describe' the concepts given. -17- typical concepts. (a list of adjectives emitted to each of the six con— cepts is included in the appendix.) Another a3pect of scale construction which can be troublesome is pairing adjectives which are functionally opposites. This Oppositeness of scale ends is a basic assumption of the semantic differential. In order in some measure to check this assumption about our scales, 21 of the more doubtful adjective pairs were presented to four Communication graduate students in an 'Antonymity' questionnaire. The presumed Opposite terms were separated and listed in random order. If the presumed opposites did not elicit each other consistently, the scale was modified to correspond to the responses obtained. Eight prOposed scales were thrown out on the basis of data Obtained in a pretest of the questionnaire; this will be discussed in more detail below. Questionnaire construction: Six comparable forms of the questionnaire were constructed, each having five different concepts judged on the same 90 scales. Concepts were assigned to questionnaires in the following manner: items within each concept category (issues, techniques, sources) were randomized, then one concept from each category in turn was drawn until five had been accumulated. These concepts then comprised a 'form,' and so on. This procedure produced inequalities between questionnaire forms of exact number of each type of concept, e.g. Form 1 received two 'techniques,' two 'sources,’ and one 'issue,‘ while Form 2 received two 'techniques,' one 'source,’ and two 'issues,’ etc. However, at least one of each type appears in each questionnaire. :Figmj:2 indicates the assignment of con- cepts to questionnaires. HHmt-lr-J Hmmr-Jr-J HHUJCDr-i HHmr-Jv-a Hmmv-lv-i HCDUJv-ir-I ~18- FIGURE TWO le-vzu—fi CONCEPTS APPEARING IN EACH QUESTIONNAIRE FORM: GROUP ONE the birth control pill 8: source concept the intrauterine device (IUD) T: technique concept sex education in schools 1: issue concept the Planned Parenthood Association Birth control GROUP TWO the rhythm method the cervical cap your family physician intentionally limiting the size of your family "Clara-... Palm] GROUP THREE sterilization of women , your friends articles on birth control in newspapers 8 magazines the population explosion teen-age parents GROUP FOUR sterilization of men legalized abortion birth control clinics the Catholic Church religious rules against birth control measures GROUP FIVE the diaphragm a pill for men social workers U. S. birth control projects in other countries state laws against distribution of birth control devices GROUP SIX sterile foams, gels, or creams the condom your parents your minister pregnancy before marriage -19- The format employed presented a concept with a single scale, then another concept with some other scale, and so on. It was hoped that in this long questionnaire (450 SD responses) the contiguity of a single concept and scale would decrease scale-scale interaction and help maintain attention, as com- pared to the more usual format in which a concept is judged on several scales in succession. The order of appearance of the five concepts was rotated mechanically, while the appearance of scales with each concept varied randomly. Five random orders of the 90 scales were generated by blindfold-draw. Scale directions were reversed randomly within each order by flipping a coin. Then each con- cept was assigned one of these random orders which determined the appearance of concept-scale pairs in the questionnaire. The aim of this randomizing was to reduce systematic bias due to item sequence or context. Presumably, the five concepts appearing in any one form of the questionnaire provide 'context' for one another which might alter judgments of individual items.* This possibility will be discussed in the following chapter. Since fatigue and boredom effects were expected to be important con- tributers of variance (30 pages, 15 SDs per page) the order of pages in questionnaires was varied mechanically during assembly. This procedure was intended to distribute the fatigue error variance evenly throughout the O questionnaire. Pretest: Ten questionnaires of Form 1 were assembled on ditto and distributed to Communication graduate students and their wives. The responses were tabulated by hand and means and standard deviations of each scale were —_A it A sample page taken from a questionnaire is included in the appendix. -20- examined. Based on this pretest data, eight of the proposed scales were discarded. A scale was deemed undesirable if (a) there were near equal numbers Of responses in each of the seven response intervals. A rectangular distribution of responses could indicate random use of the scale which would not produce correlations with other scales. (b) if the means, and actual scores, centered on the neutral response category. The scales discarded were: low brow-high brow; pious-impious; deep- shallow; lenient-strict; near-far; exciting-dull; beautiful-ugly; flexible- rigid. It is possible that some of these rejected scales would have been relevant to other concepts (or other 83) which were not included in the pretest. As well as SD reSponse data, suggestions and comments about the question- naire were solicited. Most respondents complained of the length of the questionnaire, whose average completion time was about 45 minutes, and about its apparent repetitiveness. Useful comments were also made about the instructions accompanying the questionnaire and modifications were made accordingly. The instructions used in the study are included in appendix The Respondents: Our intention was to include a wide variety of people-types without going to the length of obtaining a true random sample. Adult females were of more interest to us than others because they represent the segment of the general pOpulation Of most interest to family planning programs. Reapon- dents were drawn from several sources--undergraduates at MSU (male and female), graduate students and their wives in the Department of Communication, a large group of rural adult women (4-H leaders from many Michigan counties), and other adults from the Lansing area. The total N was 187. -21- The n for each of the six different questionnaire Forms was 31, except Forms 2 and 6 which had 32. It was found that the return rate was very low for questionnaires mailed or handed out to peOple who promised to complete them at home. Most of the questionnaires were completed by intact groups who worked in thepresence of the E. The length and nature of the questionnaire precluded personal interviewing as a means of data collection. Demographic information was collected for each respondent and since it is likely, especially for family planning concepts, that such background differences may influence judgment of concepts, this information is summarized below in terms of differences between groups. (See figures 3,4,5,6,7,8). The demographic questions and an itemization of responses are included in appendix B . It would appear that differences exist between groups with respect to sex distribution, marital status, number of children,'age, education, and residence. It is difficult to evaluate 'how much difference makes a difference' to the analyses performed in this study. Analysis: Of all questionnaires returned to us, about 30 were discarded. In most cases, this was because several pages had been left blank. Since factor analytic procedure requires complete data for each S, we could not include partially completed questionnaires. The limit allowed for missing data was the equivalent of one page (20 items) per respondent. Thus, acceptable questionnaires had to be about 96% complete. Blanks were coded u in all cases. 22 9.55 w. mow Runaway—on bouooo wanton-none 9688 Boat 2% cm. Honors II EE we EH" 23 J! .. .5. .151... - - I ll. ’ 15,11“? , FEE... t. ESSA—H; mango mu monogamous >8.o.oo 9.05. _. u 9‘ V2.1... Ill-Io as “HEAL... ..oupufl l1. 1 ult- _. he Mi: A IeIIEI Va...llr 5E ow gown“ ml mHZQHH II Ea 2h mesHo m. paououo zssoon on anHaHos mew mouoofi pououo Quezon. goat mo- ozm , .20 5% was Hug mHN bdgw 259$ ow emu—”gm: a Emoz 5 mecoo m. woooouoooao. mono wooHnonoo been. AeHHauon_nsaoowo Housoooo«.muosoo mev G. aggwg Egg “0. ...---Oi ®® @ 1 ® '|' ' }"|-A 8 ® © {AMWHMWII G .-fifi .mfifia 26 558‘ a. mafia—«Hos Hoe-H on mongoose. Soon mucosa. HmA'IIIl'I-Illllll| allllllllllllall"'ll|l'li|"'llll'|:'al: a... . o o ESE Hollillll ..I.. Isll..ll.. lljlulllii®allilal ...Iirigg! 8 ® o _ o poor o. ....................i................. ..... ......l. 1 .. 0 Eggs t. ». hum. than; N. 3. Buttons. Dukx (hm. at“ .3». ¢ok~ Factors resembling 'Qualification' occur in groups 1, 2, and 3, with scales such as COMPETENT, SUCCESSFUL, INFORMED, but also appear mixed with 'Safety' scales. The smallest factors in Mode 3 continue to reflect a concern with evaluation on a personal level and are suggestive of Berlo's 'Sociability' factor, represented by scales such as SOCIABLE, PLEASURABLE, INTIMATE, PRIVATE, INCONSPICUOUS, COMFORTABLE. However, because there just is not adequate stability of scale-use across the Mode 3 analyses for any of the factors, it remains just as easy to argue that these source-concept dimensions are versions of the dimensions appearing in other modes of the same study as to push an analogy to the Berlo dimensions. Item Reliability: In any questionnaire, and especially in a long and potentially boring one, it is important to have some check on 85' consistency in response. Such a check was provided fortuitously when it was discovered that eight scales used to judge two concepts in group #6 had been repeated on different pages of the questionnaire. Test- retest reliability coefficients were computed for these eight scales (n=32). The rtts ranged from .84 down to .37. An average reliability was computed using the z-transformation which turned out to be .68. Osgood (1, page 127) in Measurement of Meaning reports. anqu:of .85 for a much larger sample of scales and Ss. 47 Summary of Results: For Mode 1 (mixed concepts) there is discernable a reasonable and consistent pattern of factors implying a 3-dimensional meaning space for these family planning concepts. However, the correspondence between respondent groups' factor structures is by no means over- powering -- there are not a great many scales common even to four of the six groups and considering the small 'personal' factor, more intuition than science is involved in its identification. Study of coefficients of factorial similarity indicates little differentiation among a large cluster of factors and the existence of many unique factors. Despite this contradiction, there is a suggestion of a meaning- ful 3-dimensional structure strong enough to warrant further investigation. (b) In Mode 2 (technique concepts) the same factors were recognizable. The first and second magnitude factors which I have termed 'social' and 'functional' evaluation, seem more consistent, more tightly knit, than in Mode 1. That is, more scales were common to 4 or 5 groups for these two factors in Mode 2 than in Mode 1. This increase in cohesiveness of the factors is reasonable in light of the reduction in the range of concept types. However, the smallest factor, which I have called 'personal' evaluation, has become more diffuse in Mode 2. Each of the six groups produced an essentially independent selection of scales for this factor, there being only two scales common to three of the groups. Ten scales appeared twice. (see Table 5) The remaining twenty-three scales representing these factors were idiosyncratic to a group. This is not to say that the factors all 'looked like' 48 different dimensions of reSponse. Except in Groups 1 and 6, whose third factors were mixed with 'functional' evaluation items, scales appearing are consistent with a 'personal' level of evaluation. (c) In Mode 3 (source concepts) each analysis produced factors apparently unrelated to factors in other groups, as judged by scales which best represented the factors (highest loadings). Thus, it became impossible to group similar factors across groups and compile lists of common scales. Compared to Modes 1 and 2 there is clearly less commonality of response in Mode 3 across respondent/concept groups. This condition probably reflects the heterogeneous nature of the concepts categorized.§ priori as 'sources' in this study. Factor Similarity scores again Show the pattern in which the larger factors appear to be closely related to one another while smaller factors tend to be idiosyncratic. 49 CHAPTER IV CONCLUSIONS In this study we selected a variety of concepts from three areas within the tOpic area of family planhing. Scales were selected prim- arily on the basis of relevance to these concept areas. The respondents were divided into six groups, each of which judged five of the total of 30 concepts on the same 90 scales. At least one of each type of concept was included in each questionnaire form. In analysis, the groups were treated separately but in parallel fashion. In Mode I, all concepts were summed over all SS, and scale intercorrelations factor-analyzed. To this point, the study resembles a 'typical' SD factor analysis study multiplied by six. Subsequently, the same data were re-analyzed in two more modes. In MOde 2, only those concepts categorized as birth control techniques entered the factor analysis; in Mode 3, only those concepts categorized as information sources were analyzed. The result was six separate studies each with three analyses, oreighteen factor analyses in all. Darnell has said that a factor structure for several diverse con— cepts analyzed as a unit will not represent the structure applicable to any one concept, and vice-versa. That is probably especially true with the questionnaire format used in this study, i.e. C -Scale-,_C2—scale, C -scale, 1 3 etc. Aside from the effects of adding between-concept variance, it is likely that the set of (5) concepts appearing in rotation produce a unique 'context' which influences the judgment of any one concept. Thus, there may be concept-concept interactions Operating which make the resulting structure to some unknown extent unique to that one sgt_of concepts. This carries two implications: (a) if the responses to any one concept are in SO part tempered by the co-occurring concepts, then it is questionable that gx_post facto segregation of concepts for re-analysis (after the data collection) truly separates the judgments of concepts and allows independent factors to emerge...contrary to the assumption in the present study. That is, the separation of concepts in analysis is artificial, coming after the fact of judgment within a context of other concepts. (b) In the present study, a tentative conclusion is that the 'technique' concepts have dominated the 'context' of the question- naire for the reSpondents. Thus, structures primarily descriptive of technique concepts appear in Mode l and a bit more clearly in Mode 2. The apparent randomness of Mode 3 structures may result in part from the 'coalescence' of judgments toward a structure suitable for techniques. If isolating dimensions specifically relevant to information sources had been our intention, then clearly these concepts would have had to be dealt with in a 'clean' environment. Another probable cause of non-uniformity in Mode 3 (source) structures lies in the wide range of referents and experience subtended by the class of concepts called 'sources' in this study. These concepts included individuals, institutions, media; sources with which respondents will be familiar and ones with which they will not; acceptable and un- acceptable sources, and so on. The high degree of specificity of scales to source concepts probably indicates (a) that different exper- iences with and expectations about various sources make scales different— ially relevant (b) a good deal of 'denotative contamination' may be occurring, in which adjectives are used with different meanings with different concepts. Sl 'Denotative contamination', however prevalant the effect may be, is another of those intangible plagues of SD research. I do not know of any attempts to document this effect in actual data—collection situations. It would perhaps require subsidiary questions or interviews with SS to elicit their various reactions to scales across concepts. After the event Of data collection it is doubtful that instances of denotative contamination could be verified. Perhaps the notion of an effect like denotative contamination is largely the result of our wishful assumption that there is a stable and universal semantic space. If this assumption is modified to accept the presence of multiple semantic spaces in which adjectives may naturally take on different interpretations the notion of 'denotative contamination’ becomes less relevant. Denotative contamination should not be viewed as a 'fault' of the 88 but as a real part of their use of language in response to stimuli. Our measures need to utilize this phenomenon rather than reject it as error. This is another area in which we need to be able to make predictions rather than p2§£_hpg. rationalizations. A sililar kind of problem concerns SD concepts. Limiting the discussion to verbal (written) concepts representing things or people, it remains a mystery how one may determine 2_priori that a set of concepts will share the same semantic structure comfortably. There appears to be a trend away from the early Osgood notion of a really general meaning space into which any and all concepts may be fitted - toward increasing specificity of semantic spaces for certain restricted ranges 52 of concepts which meet particular research needs. In the present study, it was plain that.no single factor structure was adequate for all the concepts employed; especially in Mode 3 where it appeared that each source, or combination of two sources, generated its own unique semantic space. This effect was no doubt heightened by the presence of noise generated by group differences and 'denotative contamination.‘ The question remains...what makes concepts alike or different with respect to pertinent semantic dimensions? Osgood has said that it is impossible to work back from a location in semantic space to predict the nature of the concept (1, p. 323). This implies that there is no obvious, external, way in which concepts can be classed as belonging to the same domain of semantic space or, in fact, as being relevant to the same set of dimensions. The concepts of the present study were thought, before the fact of judgment, to be re- latively homogeneous, especially within the subclasses of 'issues,' 'techniques,' and 'sources.’ This assumption was not borne out. Turning to the 'choice of weapons' with.which to evaluate factors and factor structures: in the case of the present study, a conflict was implied to exist between an objective interpretive method (Factor Similarity coefficients) and a subjective interpretive technique which might be called 'semantic' analysis, or more frankly, intutitive analysis of the nature of factors. The usual assumption is that if there is a choice, one should opt for the most objective method available, and the more ambiguous the data, the greater the need for objective criteria of interpretation. oi 53 However, in this exploratory study, where the emphasis is more on the clue-finding than on hypothesis testing, I feel it is wasteful to reject reasonable implications - which can be tested later - in preference to impersonal but perhaps sterile con- clusions resulting from mechanical interpretive methods. At this stage, the suggestion that family planning techniques, at least, may be evaluated along three dimensions reflecting social- moral, functional effectiveness, and personal acceptance consider- ations is more heuristic and therefore more valuable than the conservative.conclusion implied by FACSIM that there is a large undifferentiated evaluation response to these concepts. The two methods utilize different information and tell us somewhat different things about factors: each has its place. 'Semantic' interpretation makes abstractions from the scales on a factor and says, 'The character of the factor is thus and so.‘ Other factors are seen as related if they can be described by the same abstraction. The input is generally the few top-loading scales which.best represent the factor. The correlational method examines every item on the two factors being compared and attempts a match between them. If the identical items do not occur with much the same magnitude the factors are reported as dissimilar. While the human observer can, somewhat perilously, equate various scales and thereby impute a similar discriminative intentions to two different sets of scales, FACSIM must respond to these as different factors. Another difficulty 54 with similarity indices occurs when dealing with 'small' factors which.do not account for a large proportion of the variance. The experimenter is left to wonder whether a low coefficient of similarity results from actual factor instability or from low loadings and few scales. The problems mentioned above suggest that a modified simil- arity index.would be useful, i.e. one which would incorporate some of the advantages of the human observer-interpreter. For instance, one might include in the computation only those scales with some minimum loading or purity on the factors being compared. Also some device allowing synonymous scales, or ones nearly so, to increase the similarity coefficient would be helpful...of course, this reintroduces the judgmental process to the comparison. The role of 'people‘ variables in SD analysis is an un- explicated area. Further work needs to be done in order that some expectations can be established about the interactions of demographic (and personality) variables with scales and concepts. It.would be reasonable to expect, for instance, that family planning concepts would.be differentially meaningful to those with marital experience and those without, to those who have children and to those who have none, to those from cities and to those from rural areas, to males and to females, and so on. In the present study, only surmises can be made about the relation.between group demographic differences and differences in factor structures, the design of the study did not permit of such tests. 55 The final.conclusion which is suggested by this study concerns the confidence one may place in any single factor— analysis based on a single Efléii group of respondents and concepts. There appears to be such a large number of uncon- trolled or unknown confounding variables at work in this research situation that generalization from some single study about 'THE' nature of semantic space for some set of concepts seems quite out of order. It must be pointed out that a problem for any correlational technique, especially factor analysis, is instability of relationships which are based on a small number of observations. It has been suggested that when psychological measures are the input to factor analysis, NS below 500 are inadequate to insure stable factor structures! Without doubt, the small group N8 of the present study (32 Ss/group) contributed to the difficulties encountered in interpretation of results. The design of the present study might be looked at as comp posed of an.'experimentall group, say any of the Mode 1 groups, and 'control1 or replication groups, having either (a) alternate questionnaire forms or (b) restricted concept-range questionnaires, whose purpose was to test the generality of conclusions stemming from the initial 'experimentalf group. We found considerable variation throughout all these ‘control' . groups, and although we could report a trend toward generality of results pertaining to technique concepts, there was enough variation among groups and sets of concepts to serve notice that generalization from any single factor analysis of SD judgments for some one set of concepts can be quite misleading. Suggestions for Future Research: The present study suggests two main areas for further research: one area concerns the applicability of 'Social,‘ 'Functional,‘ and ‘Personal' Evaluation dimensions to affective responses to family planning techniques and devices. The other area includes several questions raised by the semantic differential technique. Family planning: Replications of the Mode 2-Technique Concepts- factor analysis should be performed in which only these concepts are presented to Ss. Scales used should include those listed in Table 6, with the addition of other scales which would be likely to represent the same kinds of judgments, and a few 'anchor' scales from Osgood Activity and Potency factors. Respondent groups could be selected which closely match the populations of interest to family planning research. If a set of dimensions (and scales) can be established as highly relevant to these concepts, the resulting instrument will have many practical uses in measuring attitudes toward contraceptive techniques and in suggesting appropriate message stratagies for inducing adoption of family planning practiceS. Semantic differential research: There is more and more need for research.directed to uncovering "the principles which operate in determining a common frame of reference" for the various concepts we wish to define with the semantic differential. We need to know more about the psychologically relevant characteristics of concepts as differentiated from their linguistic or referential characteristics. 57 Until progress is made in this area, the 'absolute"3§_‘relative' question.about semantic space and the practical issue of whether one need go to the extreme of factor analyzing concepts individually, will remain troublesome. Another practical question of interest is the extent to which concepts appearing in the same questionnaire may exert mutual influence on the judgment process. That is, will responses to Concept 'A‘ (and the structure of these responses) be the same whether 'A‘ is judged in isolation or in a 'field' of other concepts? If responses to 'A' are extracted from a larger questionnaire for separate analysis, as was done in this study, how representative will these responses be? Such concept-concept context effects could be easily detected using two or three concepts (a) together (b) in isolation, and comparing the single-concept factor structures across the two situations. Denotative contamination is a notion that deserves more attention. It would be valuable to find out how widely this effect occurs and whether there are particular kinds of scales which exhibit this behavior more than others. Perhaps respondents could be interviewed immediately after - or during - SD judgment in order to map changes in the interpretation of scales across different concepts; to see which scales are in effect, functional homonyms. If other adjective scales which differentiated the various meanings of the functional homonyms. were then made available to Ss, the effects of denotative contamination might be reduced and more information obtained about the response dimensions relevant to the concepts judged. I. till lll‘lllllll all!" III. I]. ill-«Ill III! I l'l'llll ll‘ .11 in} I I 58 Finally, the relation between 'people‘ variables and differences in SD response structure should be more fully eXplicated. Q-analyses, in which the respondents are the variables forming factors, and subsequent correlation of these factors with a broad range of demographic information.might allow us to control for the error-increasing group-specific characteristics in our respondents. 1. 59 REFERENCES 03good, C. B., Suci, G. J.; Tannenbaum, P. H., The Measurement gf Meaning, University of Illinois Press, Urbana, 1957 The A. B. reader is referred to the following articles: Berlyne, D. B., Mediating Responses: A Note on Fodor's Criticisms. J, Verbal Learning §_Verba1 Behavior, 1966, 5 uOB-ull Brown, R. W., Is A Boulder Sweet or Sour? Cohtemp. Psychol,, 1958 3 113-115 Carroll, J. B., review of Measurement 2: Meaning (see reference #1 above) Language, 1959 §§_ #1 Fodor, J. A., Could Meaning Be An Rm? J. Verb. Learn. Verb. Behav., 1965 u 73-81 " Jakobovits, L. A., Mediation Theory and the 'Single-Stage' S-R Model: Different? Psychol. Review, 1966 73 376-381 Markel, N. N., The Validity Of The Semantic Differential for Psycholinguistic Analysis, J. Verb. Learn. Verb. Behav., 1966 §_ 3&8-350 -add Hunt, R. G. 6 Crapsi, L. A. to authors- Weinreich, U., Travels Through Semantic Space. Word 1958 12_ pages 3u6-366 Ross, J., Change in the Use of Semantic Differential with a Change in Context. J. Verb. Learn. Verb. Behav., 1965 g_ 1u8-151 The A. B. The A. reader is referred to the following articles: Jakobovits, L. A., Comparitive Psycholinguistics in the Study of Cultures. Internat. J, 9f_Psychol., 1966, 1, 15-37 Kumata, Hideya, A Factor Analytic Investigation of the Generality of Semantic Structures Across Two Selected Cultures. Unpublished Ph.D. Dissertation, Univ. of Illinois, 1957. Diss. Abstracts, 1958 Tanaka, Y., Cross-Cultural Compatibility of the Affective Meaning Systems. J. Social Issues, 1967, 22. 27-46 reader is referred to the following articles: Berlo, D. K., Lemert, J. B., Mertz, R. J., Dimensions for Evaluating the Acceptability of Message Sources. Report of research done under Civil Defense contract #OCD-PS-6u7l, at Mich. State University. Jakobovits, op. cit. Kumata, H. 8 Schramm, W., A Pilot Study of Cross-Cultural Meaning. Public Opinion Quarterly, 1956, 22_ 229-238 Smith, R. G., Development of a Semantic Differential for Use with Speech Related Concepts. Speech Monogrgphs, 1959, 36_ 263-272 Tanaka, Y., Oyama, T., and Osgood, C. B., A Cross-Cultural and Cross Concept Study of the Generality of Semantic Space. J. Verb. Learn. Verb. Behav., 1963 _2__ 392-n05 10. ll. 12. 13. CO Darnell, D. K., A Technique for Determining the Evaluative Discriminating Capacity and Polarity of Semantic Differential Scales for Specific Concepts. Unpublished Ph.D. dissertation, Mich. State Univ., 196a, pages 16-18, 20-21 03good, C. B., Studies on Generality of Affective Meaning Systems. American Psychologist, 1962 16' 10-28 Talbot, A. D., Generality of the Dimensions of Source Evaluation. Unpublished Ph.D. dissertation, Mich. State Unive. 1968, page 8 Darnell, op. cit., pages 18-19 Osgood, et a1., op. cit., pages 326-327 Talbot, op. cit., page 60 Sarbaugh, L. E., Generality of the Dimensions of Source Evaluation Across Language/Cultural Systems. Unpublished Ph.D. dissertation, Mich. State Univ., 1967, pages 9-10 Kiel, Donald F., Effects upon the factorial solution of rotating varying numbers of factors with differing communality estimates. Unpublished M.A. Thesis, Mich. State Univ., 1966 II. II!!! Ill’l.‘ ‘IIIIII l I I'll!" llll IIIII‘I '11) APPENDIX A: Questionnaire instruction; Sample page of questionnaire 61 INSTRUCTIONS The questions on the following pages represent an attempt to explore the ways in which people think about the topic of birth control. These are opinion questions which are designed to let you express many "shades" of feeling not possible with the ordinary kinds of questionnaires.v Here is an example of the sort of questions we will be asking: Medicare: TIMELY: : : : : : : : UNTIMELY Suppose you were asked to rate the idea of Medicare. First look at the adjectives at each end of the scale and decide which one fits best. If you felt that Medicare was 3231 timely, you would place an "X” in the space closest to TIMELY. If you felt that Medicare was yggy untimely, you would put an "X” in the space closest to UNTIMELY. The middle space along the scale stands for NO OPINION, and the other Spaces mean stronger and stronger feelings as they get closer to the ends of the scale. You might think of the spaces as being labelled like this: very quite a little neutral a little quite very /// —\\\ TIMELY UNTIMELY Some of the questions may not seem to make very much sense to you. If you find that you just cannot make a definite answer, then check the middle (no opinion) Space. This is a long questionnaire. We know it is. Your patience and cooperation are very much needed! THANK YOU First some questions about you: . . . I o A _ a v u .1 . . Ix. ,1 u r . I o) x u . .. n r . q! l "..-—.....-v...‘ .- .... .. -- 86 87 the rhythm method the cervical cap intentionally limit- ing the size of your family your family A physician large families intentionally limit- ing the size of your family your family physician large families the rhythm method the cervical cap large families the rhythm method the cervical cap intentionally limit- ing the size of your family your family physician UNTI MP LY UVTESTED : PROFOUND RIGHT RELIABLE EXACT CAUTIOUS UNJUST CONSPICUOUS INTIMATE UNTRUSTWORTHY ADVENTUROUS TASTELESS CONSPICUOUS NSOCIABLE : O. Section II TIMELY TESTED SUPFPFICIAL : WRONG UNRELIABLE INEXACT RECKLESS JUST INCONSPICUOUS REMOTE TRUSTWORTHY UNADVPVTUROUS TASTEFUL IVCOMSPICUOUS SOCIABLE c5? c53 cSu c55 c56 CS7 c58 c59 c60 c61 C63 c6u c65 c66 ’pgofi-I ;- 'l .‘ln I. l I [II ..‘I I‘ll 1“ [In III‘ ['1 II I APPENDIX E Demographic questions; Listing of demographic characteristics of respondents - by groups. 1. What is your sex? ( ( 2- What is your age? ( ( ( ( ( ( ( ( ( ( 63 Demographic Questionnaire VV VVVVVVVVVV 3. How many years of school AAAAAAA male female under 20 2o-2u 25-29 30-3u 35-39 uo-uu us-ug so-su 55-59 60 and over have you completed? o-u 5-8 9-11 12 (high school diploma) 1-3 years of college u years of college (degree) more than four years of college a. What is your marital status? AAAAA 5. How many children do you AAAAAAA G. How many children would you like to have in your family; that VVVVVVV single r-rried - living with spouse separated or divorced remarried - living with spouse widowed have? none one two three four five six or more- is, what is the ideal number to have? AAAAAAA VVVVVVV hone one two three four five six or more clS c16 cl7 clR CLS C20 9. 10. 11. For women - are you now, () () What is your religious AAA/NA VVVVV Would say you have lived INAAAA VVVVV Where would you prefer VVVVV to live if you had a choice? 6h - 2 - or have you ever been pregnant?) C21 c22 no yes.... how many times? preference? Jewish Protestant Catholic other.... please specify none ”' C23 most of your life in: cities suburbs small towns rural areas couldn't say c2u In: cities suburbs small towns rural areas not sure C25 About what income would you say your family will receive this year? AAAAAA vvvvvv In the next few years, () () () 0-2000 dollars 2-HOOO u-5,000 6-l0,000 10-15,000 more than 15,000 dollars c26 do you expect your family income to: increase stay about the same decrease c2? 65 DEMOGRAPHIC DATA 1 GROUP: VARIABLE 10 12 10 ll 21 male female- SEX 25 21 19 21 22 0-19 20-24 25-29 AGE 11 15 15 30-34 35-39 uo-uu us-ug 50-5” 55-60 60+ o-u EDUCATION 5-8 9-11 l2 13 16 1k 13 10 1-3 col. u yrs. col. 10 10 5 graduate study MARITAL STATUS single married 15 11 18 1” 12 2H 14 17 20 27 17 18 17 1H 0 # CHI LDREN l23u56 Jew RELIGION 18 21 2O 2O 23 23 Prot. Cath. other none DEMOGRAPHIC DATA - cont'd. VARIABLE GROUP: 1 2 3 u 5 6 RESIDENCE city 7 6 6 5 7 6 suburb H 7 6 10 8 8 small town 6 10 9 12 5 u rural 14 9 9 u 10 1a INCOME $0-2000 1 2 2 1 0 0 2-uooo 0 3 2 2 5 u 4-6000 5 1 6 2 u u 6-10,000 1a in 13 9 8 9 10-15,000 7 6 5 9 13 10 15,000+ u 6 3 8 l u INCOME EXPECTANCY increase 21 15 21 19 17 23 stay same 9 17 9 10 12 9 decrease 1 0 1 2 2 0 CHILD +u,5,6 u 2 5 8 7 2 PREFERENCE +3 2 8 9 6 7 u INDEX +2 3 7 3 7 8 5 +1 7 5 l 1 O 7 0 13 8 9 7 0 11 -l O 1 2 0 l 0 -2 0 l 2 2 0 2 -3 0 0 0 0 0 0 -u l O O 0 O 0 in this index, 'ideal number of children' minus 'actual number of children = child preference index. thus, the positive response categories indicate a desire for more children, a negative category may indicate S feels his family is too large. Zero = ideal matches actual number RESIDENCE +3 0 0 O O l l PREFERENCE +2 1 3 l 1 2 1 INDEX +1 3 3 6 6 M 2 O 22 18 14 16 12 22 -1 3 u 3 6 5 M -2 O 2 H 2 5 O -3 2 2 2 O 2 0 -H O O 1 O O 1 in this index, the more positive the response category, the more the S desires to live in a more urban environment than at present. negative reSponse categories indicate a desire to live in a more rural environment in the future. Zero = no residence change desired APPENDIX E: Adjective scales employed in all questionnaires - in alphabetic order. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 67 LIST OF ADJECTIVE SCALES EMPLOYED IN QUESTIONNAIRE FORMS 1-6 (in alphabetic order) adventurous-unadventurous available-unavailable broad narrow broadening-limiting careful-careless cautious-reckless 0 certain-uncertain civilized-uncivilized clean-dirty 0 clear-unclear comfortable-uncomfortab1e competent-incompetent complete-incomplete 0 constructive-destructive controlled—uncontrolled convenient-inconvenient decent-indecent decisive-indecisive definite-vague dignified-undignified discreet-indiscreet easy-difficult economical—expensive effective-ineffective 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. an. 45. 46. 47. 48. exact-inexact fair-unfair B feminine-masculine 0 friendly-unfriendly B good-bad 0 graceful-awkward 0 happy-unhappy healthy-unhealthy helpful-harmful honest-dishonest human-inhuman hygienic-unhygienic impressive-unimpressive inconspicious-conspicious informal-formal informed-uninformed B intimate-remote just-unjust B kind-cruel B,O logical-illogical moderate-immoderate modern-antiquated moral-immoral natural-unnatural 49. 50. 51. 52. 53. 5‘4. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. ADJECTIVE-LIST CONT'D necessary—unnecessary nice-nasty nontechnical-technical normal-abnormal obvious-sublte optimistic-pessimistic orderly-disorderly permanent-temporary pleasurable-painful 0 pOpular-unpopular possible-impossible practical-impractical private-public 0 profound-superficial proper-improPer pure-impure qualified—unqualified B reliable-unreliable respected—unrespected reSponsible-irresponsible right-wrong safe-dangerous B sane-insane 0 scientific-unscientific 0 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88. 89. 90. serious-humorous simple-complex sociable-unsociable soothing-disturbing sophisticated-naive O 0 13,0 0 successful-unsuccessful superior-inferior sympathetic-unsympathetic tasteful-tasteless tested-untested timely-untimely thoughtful-thoughtless tolerant—intolerant O trustworthy-untrustworthy useful-useless virtuous-sinful wise-foolish young—old O Scales marked with an '0' appear in the Osgood Thesaurus study. marked with a 'B' appear in the Berlo et a1 Lansing study. Scales O APPENDIX D: Listing of Scales Selected by Factor-loading to Represent Each Factor; 3-Factor Solutions; Modes 1-3. O \ \C‘ MODE I, GROUP 1 Concepts: pill, IUD, sex education, PPA, birth control "SOCIAL" Scale Purity Loading logical- .94 .81 clean- .86 .82 right- .88 .79 fair- .83 .79 hygienic- .94 .78 moral- .91 .78 sane- .87 .78 honest— .85 .78 human- .94 .77 good- .90 .77 decent- .85 .77 kind- .86 .76 proper- .79 .76 virtuous- .82 ' .75 constructive- .77 .74 pure- .89 .73 civilized .83 .73 just- .82 .73 "FUNCTIONAL" definite- .91 .77 competent— .83 p .75 exact- .86 .73 easy- .84 .72 certain- .91 .72 complete- .99 .70 clear- .86 .70 graceful- .84 .69 successful- .75 .67 effective- .76 .60 reliable- .74 .60 "PERSONAL" intimate- .71 .55 natural- .60 -.48 private- .81 .47 inconspicuous .99 .46 obvious .99 -.4O MODE 2, GROUP 1 Conceptsz= pill, IUD "SOCIAL" Scale kind- human- fair- clean- thoughtful- sane- good- logical- decent- constructive- just- moral- happy- hygienic— proper- tolerant- wise- pure- honest- impressive— "FUNCTIONAL" complete- natural- certain- safe- sociable— simple- pOpular- clear- tested- soothing- "PERSONAL" discreet- broad- controlled- SOphisticated- qualified- obvious- optimistic- Purity .83 .82 .77 .80 .84 .82 .78 .77 .76 .75 .82 .75 .74 .73 .66 .88 .74 .73 .65 .80 .84 .93 .85 .78 .74 .99 .77 .65 .87 .64 .69 .97 .83 .63 .58 .94 .57 Loading .81 .81 .81 .80 .78 .79 .79 .78 .76 .76 .74 .73 .72 .72 .70 .69 .69 .69 .69 .68 .73 .71 .69 .67 .65 .64 .64 .61 .59 .58 .66 .65 .65 .61 .57 .55 .55 MODE 3, GROUP 1 Concepts: sex education, planned parenthood Scale Purity right- .89 moral- .85 possible- .88 necessary- .75 logical- .91 hygienic- .80 virtuous- .71 constructive- .64 proper- .72 decent- .70 safe- .71 normal- .65 good- . .85 practical- .82 tasteful- .72 successful- .90 respected- .80 trustworthy- .79 exact- .75 definite- .84 competent- .75 superior- .70 graceful- .65 orderly- .58 certain- .88 sociable- .79 moderate- .73 pleasurable- .71 happy- . 6 3 intimate- .97 obvious- .89 modern- .71 Loading .83 .78 .77 .76 .75 .74 .74 .74 .72 .71 .68 .67 .66 .66 .66 .79 .75 .74 .74 .73 .73 .70 .69 .69 .68 -.66 -.64 -.60 -.59 -.57 -.51 -.50 \J {\3 MODE I, GROUP 2 Concepts: rhythm method., cervical cap, physician, intentionally limiting family, large families "FUNCTIONAL” Scale Purity Loading trustworthy- .81 .81 exact- .89 .80 certain- .82 .80 reliable- .76 .78 competent- .82 .76 practical- .81 .76 effective— .84 .75 definite- .79 .71 successful- .77 .71 modern- .74 .71 logical- .67 .71 scientific- .72 .69 cautious- .72 .67 "PERSONAL" economical- .88 .64 inconspicuous .96 .49 private- .96 .48 informal- .69 .46 serious— .73 .45 "SOCIAL" friendly- .86 .74 nice- .83 .74 happy- .72 .74 natural- .78 .72 normal- .75 .69 comfortable- .71 .68 sympathetic- .69 .67 sociable- .90 .66 honest- .67 .65 virtuous- .65 .65 MODE 2, GROUP 2 Concepts: rhythm "FUNCTIONAL" Scale trustworthy- exact- reliable- successful- competent- effective- certain- practical- definite- scientific- responsible- superior- "SOCIAL" moral- virtuous- civilized- serious- respected- dignified- hygienic- healthy- pure- broad- useful- sane- clean- honest- discreet- "PERSONAL" pleasurable- nice- convient- comfortable- friendly- happy- soothing— easy- intimate- sociable- method, cervical cap Puripy .90 .89 .85 .85 .89 .84 .82 .81 .83 .88 .82 .65 .88 .94 .84 .92 .68 .70 .83 .58 .66 .57 .62 .58 .87 .76 .71 .84 .80 .67 .78 .64 .63 .66 .71 .87 .69 Loading .89 .83 .81 .81 .80 .80 .79 .75 .74 .72 .70 .68 .81 .80 .77 .76 .68 .67 .65 .65 .60 .60 .59 .58 .57 .57 .57 .76 .70 .70 .69 .68 .68 .63 .60 .55 .55 MODE 3, GROUP 2 Concepts: your physician Scale careful- reSponsible- useful- orderly- scientific- informed- pOpular- sympathetic- modern- healthy- civilized- friendly- effective- logical- helpful- thoughtful- clean- natural- exact- impressive- optimistic- SOphisticated- obvious- intimate- inconspicuous- sane- informal- graceful- prOper- dignified- pleasurable- easy- qualified- possible- tolerant- broadening- adventurous- PuriEy .81 .81 .79 .99 .95 .89 .86 .80 .93 .80 .87 .72 .75 .68 .78 .67 .66 .65 .83 .68 .87 .93 .81 .85 .80 .57 .95 .91 .83 .76 .75 .96 .64 .59 .58 .60 .83 r~v ’4 Loading .87 .86 .85 .84 .84 .84 .83 .83 .82 .82 .81 .81 .79 .79 .78 .78 .78 .78 .77 .77 .84 .71 .66 .64 .52 .48 .84 .82 .78 .77 .74 .72 .68 .67 .66 .63 75 MODE I, GROUP 3 Concepts: steriliz. women, your friends, articles on B.C., pop explosion, teen parents "SOCIAL" Scale Purity Loading kind- .94 .88 good— .88 .87 helpful— .86 .85 nice- .96 .83 proper- .87 .83 constructive- .87 .82 decent- .87 .82 pleasurable- .96 .81 friendly- .94 .81 healthy- .88 .80 comfortable- .83 .80 sociable- .96 .79 just- .82 .79 fair- .78 .79 tolerant- .83 .78 wise- .76 .78 "FUNCTIONAL" decisive- .92 .78 definite- .86 .75 certain- .97 .74 clear— .80 .67 broad- .70 .67 effective- .83 .66 tested— .84 .59 reliable- .65 .64 exact- .71 .59 successful- .59 .62 "PERSONAL" inconspicuous— .82 —.62 obvious- .96 .61 private- .90 -.52 young- .99 .45 informal- .81 .45 MODE 2, GROUP 3 Concepts: sterilization of women "SOCIAL" Scale Purity thoughtful- .94 nice- .93 good— .89 right- .91 prOper- .89 cautious- .87 just- .91 fair— .87 pure- .97 necessary- .94 helpful- .87 tolerant- .99 logical- .88 kind- .86 wise- .77 qualified- .94 constructive- .78 sociable- .88 sane- .78 graceful- .82 "FUNCTIONAL" effective- .93 competent- .82 decisive- .97 reliable- .88 successful- .84 certain- .83 complete- .90 tested- .88 exact- .94 safe- .70 useful- .64 possible- .70 "PERSONAL" honest- .48 obvious- .71 non-technical- .63 informal- .85 adventurous- .76 Loading .90 .89 .88 .87 .87 .87 .86 .85 .84 .84 .84 .83 .83 .83 .82 .81 .81 .80 .80 .79 .83 .81 .76 .76 .76 .76 .73 .73 .72 .70 .69 .67 .62 .59 .53 .52 .51 .l'lil‘llllllllllllllll'lll MODE 3, GROUP 3 your Concepts: Scale civilized- orderly- responsible- safe— decent- prOper- careful- kind- good- sane- soothing- comfortable- nice- trustworthy- fair— logical- constructive- competent- broad- thoughtful- exact- successful- reliable- decisive- modern- certain- healthy- useful- clear- natural- informed- serious- private- informal- friends, articles on 8.0. in newspapers and magazines Puripy .88 .85 .84 .86 .83 .87 .86 .77 .81 .80 .88 .87 .87 .81 .89 .88 .79 .71 .79 .74 .74 .70 .68 .68 .89 .78 .78 .74 .99 .83 .95 .94 .98 .83 77 Loading .82 .82 .82 .81 .79 .78 .78 .77 .76 .76 .75 .74 .74 .74 .74 .82 .81 .78 .77 .76 .74 .73 .73 .73 .71 .71 .70 .7O .69 .69 .78 .75 .58 .50 MODE I, GROUP 4 Concepts: steriliz. women, legalized abort., B.C. clinics, Cath. church, religious rules "PERSONAL" Scale Puripy Loading modern- .92 .82 young- .99 .76 scientific- .87 .76 technical- .87 .72 timely- .68 .69 broad- .63 .68 practical- .60 .69 adventurous- .96 .66 possible- .72 .66 "SOCIAL" moral- .97 .82 decent- 1 .88 .79 human- .80 .79 dignified- .33 .78 nice- .74 .78 pure- .88 .77 friendly- .75 .77 good— .73 .77 proper- .73 .77 virtuous- .89 .76 respected- .82 .76 kind— .74 .76 constructive- .68 .76 tasteful— .73 .75 healthy- .71 .75 honest- .78 .74 "FUNCTIONAL" definite- .91 -.69 exact- .81 -.67 decisive- .76 -.66 clear- .90 -.65 certain- .76 -.62 profound- .66 -.55 permanent- .79 -.49 obvious- .91 -.45 MODE 2, GROUP 4 Concepts: steriliz. of men, legalized abortion "SOCIAL" Scale just- nice- civilized- kind- right- human- moral- soothing- dignified- constructive- wise- decent- fair- good- prOper— sane- thoughtful- tasteful- friendly- necessary- healthy- logical- "PERSONAL" inconsPicuous- modern- adventurous- non-technical- permanent- young- serious— "FUNCTIONAL" reliable- effective- successful- exact- profound- definite- scientific- decisive- certain- intimate— Purity .93 .93 .98 .94 .91 .89 .96 .94 .91 .91 .88 .96 .88 .82 .82 .93 .91 .87 .92 .90 .84 .84 .98 .86 .87 .50 .66 .88 .87 .84 .80 .61 .95 .61 .66 .63 .94 .73 .74 Loading .91 .90 .89 .88 .88 .87 .86 .86 .86 .86 .86 .85 .85 .85 .85 .84 .84 .84 .83 .82 .82 .82 .67 .67 .54 .54 .53 .49 .48 .74 .69 .67 .64 .60 .58 .56 .55 .55 .53 MODE 3, GROUP 4 Concepts: B.C. clinics, Cath. Church Scale modern- tolerant- practical- scientific- broad- logical- young- thoughtful- sympathetic- fair— adventurous- informed- right- timely- wise- moral- virtuous- respected- dignified- prOper- honest- decent- reliable- kind- hygienic- trustworthy- careful- definite- clear- serious- permanent- controlled- obvious- PuriEy .94 .83 .83 .95 .81 .77 .97 .84 .76 .72 .95 .74 .64 .77 .64 .89 .92 .80 .81 .69 .70 .68 .70 .64 .62 .59 .69 .87 .60 .87 .82 .76 .98 Loading .90 .85 .85 .83 .83 .82 .79 .79 .79 .79 .78 .75 .75 .73 .73 .85 .81 .79 .78 .74 .73 .68 .68 .67 .67 .67 .65 .77 .67 .64 .59 .57 .51 81 MODE I, GROUP 5 Concepts: diaphragm, pill for men, social workers, U.S. B.C. projects, state laws re-B.C. "SOCIAL" Scale Puripy Loading civilized- .97 .91 fair- .99 .89 constructive- .98 .89 human- .98 .89 just- .96 .89 decent— .96 .88 kind- .97 .87 nice- .98 .85 thoughtful- .94 .85 wise- .90 .85 sane- .90 .85 right— .92 .84 timely- .94 .83 moral- .93 .83 proper- .93 .83 logical- .94 .82 necessary- .97 .81 healthy- .95 .81 sympathetic- .92 .80 tolerant- .88 .79 "PERSONAL" tested— .90 -.65 available- .90 -.54 feminine- .96 -.50 "FUNCTIONAL" certain- .77 .62 intimate- .80 .55 complete- .72 .54 exact- .64 .54 easy- .90 .45 definite- .59 -.53 MODE 2, GROUP 5 Concepts: diaphragm, pill for men "SOCIAL" Scale civilized- just- virtuous- proper- decent- sympathetic- moral- nice- constructive- kind- qualified- moderate- tolerant- pure- cautious- fair- possible- logical- "PERSONAL" modern- graceful- convenient- tested- feminine- scientific- broadening- young- pleasurable- comfortable— "FUNCTIONAL" certain- complete— successful- helpful- exact- reliable- competent- trustworthy- safe- intimate- Puripy .96 .94 .97 .95 .92 .99 .94 .85 .85 .84 .88 .99 .93 .88 .78 .84 .98 .84 .94 .76 .75 .63 .86 .77 .65 .95 .60 .59 .99 .99 .77 .72 .91 .86 .60 .66 .57 .97 Loading .90 .88 .87 .87 .85 .84 .83 .81 .80 .78 .76 .75 .75 .75 .75 .74 .71 .71 .72 .70 .67 .61 .57 .57 .57 .55 .54 .50 .73 .72 .72 .69 .68 .67 .66 .63 .63 .59 ‘II , I III ll).|[ 1 ..... MODE 3, GROUP 5 Concepts: social Scale timely- nice- human- useful- necessary- civilized- constructive- thoughtful- good- wise- superior- tolerant- dignified- hygienic- virtuous- non-technical- moderate- controlled- decisive- successful- popular- logical- comfortable— Optimistic- convenient- easy- happy- broadening- workers Purity .81 .78 .78 .71 .82 .69 .69 .67 .70 .68 .78 .65 .88 .72 .73 .95 .61 .75 .73 .85 .73 .59 .67 .90 .74 .99 .65 .90 “cf.“ 3 Loading .84 .80 .80 .77 .76 .76 .76 .76 .74 .74 .73 .71 .83 .68 .66 .65 .65 .63 .59 .79 .75 .71 .70 .67 .66 .63 .60 .55 MODE I, GROUP 6 Concepts: PB4M "SOCIAL" Scale honest- proper- decent— kind- wise- moral- good- reSponsible— virtuous- right- constructive- necessary- safe- respected— useful- pure- dignified- civilized- sane- thoughtful- "FUNCTIONAL" definite- intimate- modern- decisive- effective- broadening- profound— "PERSONAL" private- feminine- serious- obvious- informal- inconspicuous- fms., gels, crms.; PuriEy .97 .97 .94 .93 .92 .99 .93 .95 .98 .95 .93 .96 .93 .90 .94 .94 .89 .88 .89 .86 .75 .95 .73 .63 .56 .68 .55 .99 .74 .56 .70 .55 .57 ‘84 condom; your parents, Loading .90 .90 .90 .90 .90 .89 .89 .88 .87 .87 .86 .84 .84 .84 .83 .83 .83 .83 .82 .82 .62 .59 .59 .58 .58 .57 .53 .56 .46 .45 .44 .42 .40 your minister; [iil‘llllll [.1 ll 1 '11 MODE 2, GROUP 6 Concepts: foams, gels, creams; "SOCIAL" Scale right- wise- kind- proper- hygienic- decent- healthy- constructive- civilized— good- clean- moral- reSponsible- logical- virtuous- necessary- honest- thoughtful- "PERSONAL" braodening- graceful- braod- convenient- superior- profound- modern- soothing- decisive- exact- "FUNCTIONAL" tested- trustworthy- safe- available- Purity .94 .91 .96 .93 .81 .88 .89 .97 .82 .81 .89 .84 .86 .73 .81 .90 .78 .64 .96 .96 .87 .99 .62 .59 .75 .82 .58 .53 .92 .97 .49 .92 condom Loading .87 .85 .84 .83 .82 .81 .80 .79 .78 .78 .76 .76 .75 .75 .74 .71 .71 .7O .74 .71 .71 .70 .66 .66 .65 .63 .63 .63 .66 .55 .54 .51 MODE 3, GROUP 6 Concepts: your Scale decent- careful- virtuous- proper- wise- good- pure- convenient- superior- exact- SOphisticated- modern- broadening- young- inconspicuous- serious- adventurous- kind- natural- just- tolerant- friendly- fair- normal- sane- pleasurable- happy- possible- helpful- right- soothing- successful- C"). 0\ parents, your minister Purizy .80 .91 .86 .82 .69 .72 .88 .87 .71 .94 .95 .84 .86 .78 .94 .72 .71 .78 .79 .85 .90 .73 .72 .69 .86 .92 .89 .85 .67 .81 .95 .72 Loading .78 .75 .75 .75 .73 .72 .70 .66 .66 .62 .74 .68 .66 .54 .51 .49 .49 .81 .76 .74 .72 .72 .72 .72 .71 .71 .70 .70 .70 .69 .64 .62 11!. ' || 1 I. l I' I 5 II I l 'l _ ll [ill 1" I .11 lllfl 1 I I I] | l l APPENDIX E FACTORIAL SIMILARITY TABLES: PART 1 -- Check of Predictions of Factor Similarity Based on Observation of Scales With Highest Factor Purity; Across All Modes; 3-Factor Solutions. a - factors thought to be 'Social' Evaluation factors compared b - factors thought the be 'Functional‘ Evaluation factors compared c-- factors thought to be 'Personal' Evaluation factors compared PART 2 -- Comparison of All Factors Within Each Mode of Analysis; 3-Factor Solutions. a - all Mode 1 factors compared b - all Mode 2 factors compared ~c - all Mode 3 factors compared PART 3 -- Check of Predictions of Factor Similarity Based on Observation of Highest Loaded Scales; Across Modes 1 and 2. a - factors thought to be 'Social' Evaluation factors compared b - Factors thought to be 'Functional' Evaluation compared c - factors thought to be 'Personal’ Evaluation compared Abbreviations -- G1, G2, G3, etc. refers to Group number 1-6 F1, F2, F3 refers to order of appearance of a factor within any 3-factor solution. Is independent of the 'type' of factor. Mode 1, 2, 3 refers to type of concept included in the factor analysis. Mode 1 included all five concepts from each questionnaire. Mode 2 included only 'technique' concepts from each questionnaire. Mode 3 included only 'information source' concepts from each questionnaire. Significance of PAC-SIM coefficients: a guideline to interpreting the 'significance' of these scores is given by the 'lower limit of best fit' develOped by Dr. Charles Wrigley at Michigan State University. For 3-factor solutions, this value is approximately .79. We may therefore consider that any relationship of .79 or above in the following tables indicates a 'significant' similarity of factors. (decimal points are omitted in the tables) 1 I! lllll'llll' 1". | 'l .IIIIIIIIII II II) I III III I... l I [II ’3‘ mIHmu A mmoaoom groode HO we .moowmw. mHH zone H mendoom noewmdea moose H moose M macaw w moose : H\H H\M H\w M\H M\M M\w w\H w\M w\w :\H c\M :\w 0H WH WM Qm Wm om HM OM WH mm» mos Ho WM QQ mm Mm mm mm no» mo» OM mm» mH mm wH mos MH» om mt» mm mm» WM MM» 00» om Mo» mo MH» Qm Mm 0Q 0Q m: HH M: om om oo 0: WH m5» QQ MH mm» mo Qm was QM» om WM ms» mm» o: was me MM“.x mm» mm» om MHs Wm MMs mQen om was mM Qw mm mm» om ow Q: mm WH MQs mH» om mQ» QM MH» cw» mm» om mm% mm» Qw WM :Q mo om mM :Q :H :w mQ om Mm to mm Wm mo mo» Mw mm» mH Qm QQ m:» H: QQ Q: mHs om WH was m:» CH mm» Qw we» ems mme om MH» 0Q» Qm WM Qo Qm Hm QQ :M Q: mQ mo» 00 Q: QM QM mm so :M cw cm mm MQ mm mm mm mm wM wm 0896 m m\H m\M to Q: :M mm» so Qm as mm mm 080cm m m\w m\H m\M m\w Qm Q: mm :m mo Mm mnMo >HH.zooe M mmOflOdm nosomden moose H macaw M atone m oeocw : adoco m meoco m H\H H\M H\w M\H M\M M\w w\H w\M w\w :\H :\M :\m w\H m\M m\m m\H m\M m\w DH WH WM Q: mm QM mm QM WH QQ mM Qm wM mQ» QQ Qm m » mm mm» Qm m: mQ QM Om WH mm» MH» QH Qw mm» mm» WM QM mt» Qw ms» Qm mm mm Mm mm om w: Hm wM HQ Mw om a: mH MM» MMm QM QQ mo» mm» mm» Qw MQ WM mm OM w: MQ Mm HQ MM ow mm Mm Wm Qm mm» Qm MQs Q; QH Qw mm» Mw QQ Mm mm mH 0H» mos Qm» Qw MM» Qm mm» QM we was Mm. Qm wM mm to cm mo :Q m: mo wQ 00 mM :H cw cm mm Qm sz QH_ was Qm QM Qw sz Hm Qm HH mms Q0 Q0 om WH MHs Qm Qm Qw MHs QQ mos Qm w: MM» Mm QQ MMs mo Qm MM Qm Qm Qm mH» QM mos Qm Qm HM Qm MQ mos QM mM QM QM mm mM Qw Qm Qm mm :0 mo Qw Hm mm Ho QM mm Mm Qw mM m: e i III-”a - .v a. 1 .s 93 WIMO >HH some M WeOMOUm ooewmoea Moocw H MdOCW M MoOCW M Mdoco : MdOCW M MGOCW M H\H H\M H\M M\H M\M M\M M\H M\M M\M :\H :\M :\M M\H M\M M\M M\H M\M M\M OH WH WM MM WM QH MM MM WH Mo“.fl MM“.n QMs WM Mo» :Q :5 :Q WM MMr MH“.n QM MM» :M OM WH MQs QQ QM» MM» :M MQs WM QM MMe MM Mrs MM M s QM WM OM HH HH HM o: H: HH HO O: WH MMs QMs QMM MMs Mo Mrs MMs MHs 0Q WM MMs MMs QMM MHs .MH MMe MMs Mow or MH% WM QH MO QMm QMw Mo QQ QH QM OM M: QM OM WH MMs QM QM» MHs :M MMM MQM MMM Ho MMs MQ» QM» WM QM QM» QM Mos :M QM QM QH OM Q0 MM» Q s QM» WM Q » QM QM M s :Q QM MMs QQ OM MMs MMs QM QM MM OM WH MMs QM QM Mo» :H MM» MQ".u QM OH Q: MMs QM Mrs MMs QM WM :M Mo MH MH MM MM 3M Qo HM MM MM :5 MM M: MM :Q WM MMs QM MM MM» M: MMs Mos QQ HM MM» MQ» QM MM“.fl MM MM» QM MH Mode 1 Mode 2 Mode 1 Mode 2 man WmOMOHm asocmrd Ho MH WH MM WM MM WH 0: WM MM WH MM WH MH WH MM WM MM WH m: WH MM WH MM WH MuMo zone H H\H M\M MOM MOM MMM M M MMM MQM MHM MwM MOM MQM MMM MHM mwM MOM MMM MMM MHM MMM MMM we» was WmOMOdm asocmra 40 DH MM MM m: MM MM MH MM MM a; MM MM WM WH WM WM WM WM WM WH WM WM WM WM MOM MOM MQM MOM QM MM.fl MOM MMe MMs MOM QH MOM MMM was QQ MmM MMM MQM MMM mmM QM we .moonH. m