A TECHNIQI]N FOR THE OF ATTITT]DES MEAST]REMENT BY REI{SIS LTT(F'B1, Ph.D. NEW YORK UNIVERSITY ARCHIVES OF PSYCHOLOGY R. S. WOODIYORTE, Eorron No. 140 nI firch;npt ?syrL.l5y zz (n5z -3s) Vr(v^e ((' s-s5 NEW YOITK i June, 1932 \ i TABLE OF CONTENTS Page o II. Problem III. Procedure 11 Results IY. 2L . 1. The Sigma Method of Seoring. . 2L . 2. The Simpler Method of Scoring. . 25 . 3. The Reliability of the Scales. . 28 Scales 4. The Validity of the 32 5. Comparison of the Simpler Method with . the Thurstone Method of Scoring. . 33 V. Psychological Interpretation YI. Summary 42 Appendix, The Method of Constructing an Attitude Scale 44 Bibliography 54 A Technique for the Measurement of Attitudes I. INTRODUCTION* Attempts to measure the traits of character and personality are nearly as old as techniques for the measurement of intel- lectual capacity, yet it can scarcely be claimed that they have achieved a similar suecess. Part, at least, of the difficulty has lain in the statistical difficulties which are encountered when everyday aspects of soeial behavior, ordinarily handled as qualitative affairs, are treated from the mathematical point of view. The present study, although part of a larger in- vestigation undertaken in L929 by Gardner Murphy, aims primarily at the solution of a technical problem which has arisen in relation to the quantitative aspects of the study of social attitudes. The history and present status of research upon personality traits in general, and social attitudes in particular, have been so thoroughly surveyed by Murphy (21, pp. 381-386, and 22, pp.558-690), Bain (4), Vetter (41), Katz and Allport (16), W'atson (43), and others, that no useful purpose would be served in attempting such a study here. Nevertheless, among the hundreds of efforts to measure social attitudes during the Iast few years, the careful pro- cedures developed by Thurstone (34, 38) have naturally and rightly received speeial attention. These are characterized by a special endeavor to equalize the step-intervals from one attitude to the next in the attitude scale, using the familiar methods of psychophysics for such determinations. The Thurs- tone methods have been shown to yield a satisfactory relia- bility, and, in terms of correlations between scores and case histories as evaluated by judges, a satisfactory validity (Zg). Many obvious affinities appear between the present study and those of Thurstone, yet in a sense the present report _ * This study is one of a series under the general direction of Professor Gardner Murphy. 5 constitutes a radical departure from the concepts which Thurs- tone has published, as, for example, in the use of judges. A number of statistical assumptions are made in the ap- plication of his attitude scales,-e.g., that the scale values of the statements are independent of the attitude distribution of the readers who sort the statements (38, p. 92),-assump- tions which, as Thurstone points out, have not been verified. The method is, moreover, exceedingly laborious. It seems legitimate to inquire whether it actually does its work better than the simpler scales which may be employed, and in the same breath to ask also whether it is not possible to construct equally reliable scales without making unnecessary statistical assumptionis. Since so much is being published about attitude measurement, it seems worth while to raise these questions and to report on some results relative to the problem. It is feared that some will mistakenly interpret this article as an il\ "attack" on Thurstone's methods. I therefore wish to em- phasize in the strongest terms that I am simply endeavoring to call attention to certain problems of method, and that I am very far from convinced that the present data close the ,I question. 'j '1 ,il i' rich Thurs- rdges. II. PROBLEM in the ap- rale values From an historical point of view one fact deserves emphasis, istribution namely the extraordinary interest in recent years in the prob- i.* of the "gererality" or "specificity" of character traits-a :-na svseurimfiepd-. problem which in its simplest form has to do with the question It of ttr. functional independence of social habits. Roughly seems the speciflcity theory would define personality as 'ork better speaking, rnd in the ,-.o*po.ite of many independent habits, while those who em- phasize the relative unity of character point to signiflcant l construct general factors which make for effective prediction from be' statistical havior in one situation to behavior in a different situation. ut attitude It to the present writer that the problem of specificity questions seems em. It has been greatly misunderstood by many writers upon social is attitudes and that the value of such contributions as the pres- ;icle as an sh to em- ent may lie chiefly in redirecting attention to those aspects of the speciflcity-generality problem upon which quantitative rdeavoring ,nd that I results are clear-cut and psychologically signiflcant. It will , close the be well first to consider briefly the logic of the specificity-gen- erality problem, then to present and evaluate our data, and finally to undertake to flnd the psychological significance of our data in relation to the problem raised. If an attitude be defined by the social psychologist as a tendency toward a particular response in a particular situa- \ tion, it is clear that the number of deflnable attitudes existing in a given person at a given time will depend upon the range of stimuli to which he is subjected. But since it is possible to group stimuli in almost any conceivable manner and to elassify and subclassify them indefinitely, it is strictly true that the number of attitudes which any given person possesses is almost infinite. This result is statistically as well as psycho- logically absurd. Exactly the same absurdity and the same obstacle to research is offered by those definitions of attitude which conceive them merely as aerbal erpressions or as in- dications of assent to or dissent from particular verbal expres- sions. The number of possible verbal combinations is, of course, infinite and the number of attitudes must on this basis likewise be so. It is clear then that those who have defined attitudes in the above two r,vays have not meant exactly what 8 A TECHNIQUE FOR THE they have said. They have really intended to indicate not the actually discriminable tendencies to oVert action or the verbal- response patterns but certain discernible groups of social responses. Within each group a family resemblance of the various responses is assumed, and each group of attitudes is supposed to show some distinguishable difference from every other group. If the analysis is pressed far enough, this turns out to mean that the attitude is a habit sufficiently compact and stable to be treated as a unit. It will, of course, be recog- nized that variations of an individual's response within this sphere of a given "attitude," together with differences betw'een each attitude and the next, are involved. If it is my "attitude" to regard the eating of starches as a dietetic monstrosity, either an inherited dislike for such foods or a bundle of ac- quired tendencies directed towards bread, potatoes, rice, etc., must be assumed to exist. Whether we take the attitude therefore as an entity innate or learned, it is in either case not an inflexible and rigid element in personality (if, in fact, any . such elements exist), but rather a certain raruge witltin whiclt rl )esponses lnoae. On this basis one of our cardinal problems is to find whether social attitudes, in this sense, can be shown to be measurable, and if an affirmative answer is forthcoming, a serious attempt must be made to justify the separation of one attitude from if others. For there are no family differenees between atti- tudes, there is simply one infinite series of attitudes. On this basis the measurement of attitudes could never mean any- thing more than the determination of an amount of some one tendeney present, but what this one tendency aetually is could never be defined. Perhaps vaguely realizing this difficulty, many investigators in the fleld of attitude measurement have assumed a sharp distinction between different families of attitudes, and per- fect resemblance or indeed identity between the members of any given family of responses. Take, for example, the fre- quent raising of questions as to whether there is a relation between internationalism and economic liberalism. Such termi- nology is based upon two assumptions: first, that there is an entity to be called internationalism, the boundaries of which i can be so defined as to separate it from economic liberalism and permit one to tell which is which-otherwise the state- i MEASUREMENT OF ATTITUDES it rte not the ment is tautological-, second, is assumed that witluin the ;he verbal- sphere of internationalistic responses there is good prediction of social from one response to the next. Internationalism is treated rce of the as an entity such that the defining of a person's international- ttitudes is ism on a given issue would deflne his internationalism on all rom every others. this turns We have not been concerned up to the present, to define the y compact term "attitude," since the logic of the above argument seems , be recog- to be the same when applied to either of two prevailing defi- irithin this nitions of attitude. Contemporary definitions cluster about :s betwben two chief conceptions: first, that_attitUdeg-_a_rg dispositions "attitude" ioward-overt aglion ; second, thfthey uru *rUat if^6-sTifutes onstrosity, for overt action. The former usage seems to the present rdle of ac- writer to be preferable.* The verbal declarations of opinion , rice, etc., and attitude are regarded as an indirect method of measuri,ng re attitude dispositiozs which are most easily signifled and expressed in gr case not verbal form. It is desirable, however, to point out that the r fact, any same considerations regarding specificity and generality apply thi,n whiclt, to the verbal declarations as to other social habits. The same problem of group factors and special factors appears even rd whether when dealing with paper and pencil behavior. ' reasurable, consequently whether attitudes be defined in terms of un- us attempt derlying dispositions toward overt action or in terms of mere Itude from verbal declarations, investigators have proceeded almost with- ;ween atti- out question upon the assumption that social attitudes are s. On this grouped in patterns or clusters. A series of verbal proposi- -r mean any- tions dealing with the same general social issue are assumed f some one to be more or less equivalent, or at least to be closely related lly is could so as to permit prediction from a knowredge of a subject,s attitude on one issue to the same subject,s attitudes o, oth." vestigators aspects of the same issue. similarly, overt behavior favorable rr rd a sharp to or in opposition to a racial, national, religious, or economie L l, and per- group, is assumed to have some predictive value in relation to rembers of conduct in the future when the sa.me group or class is again le, the fre- involved. Speaking generally, attitudes may be conceived as.r a relation "clustered" or linked together; a general pro-Japanese atti-\ iuch termi- tude, for example, may show itself in a series of pro-Japanese there is an s of which x If an exact definition be. regarded as useful, the following (Century E liberalism Dictionary and Cgclopedia, 1904-1906, I, 974) ii preferred: n'reiation oi , the state- .$T' persons viewed as the expression of, or as affecting, feeling, opinion,- in- tentions, etc.', ffit $, rl 10 verbal declarations or a series of pro-Japanese overt acts. In statistical language, a group factor is assumed at the outset. Now just as the student of mental organization recognizes group factors which vary greatly in their magnitude and signiflcance, so the student of attitudes tends to think of rela- tively important group factors occurring in some attitude areas and relatively unimportant ones in others. Here, how- ever, the agreement among the investigators ends. In the few short years during which any serious statistical sophis- tication has existed in attitude research, much acrimony has been evident between those who are concerned to regard the group factors as slight and unimportant, and those who regard them as large and signiflcant. Thus we flnd Bain (4) insist- ing that attitudes cluster together only when they constitute alternative verbalizations of the same conative tenclency in relation to a speciflcally deflned social situation. The various studies in the speciflcity of character traits have led to a wide- spread and, it must be admitted, highly uncritical acceptance of the view that verbal propositions have no predictive value for anything else in life except similar verbal propositions confronted by the same subject under the sar,e -ircumstances. On the other hand, an excessively general view of character traits is naively assumed with almost equal frequency by writers who have discussed radicalism and conservatism, some of whom have deflned the characteristics of radicals and con- servatives, suggesting evidence as to the psychological mech- anisms making for such general trends. Thus as early as 1924, H. T. Moore (20) deflned flve basic psychological char- acteristics predisposing towards radicalism as contrasted with conservatism, and as recently as 1929 and 1930 G. W. All- port (2, 3) has argued for a cluster of traits, chiefly intellec- tual, which make for radicalism on political, economic, and other social issues. It seems to be hish time to cut through the statistical con- fusion which has resulted from the whole speciflcity-generaiity argument, to bring out the actual points of disagreement which separate these rival groups of psychologists, to make clear the statistical assumptions involved in all such methods of reasoning, and above all to test empirically in an extensive way the actual coherence or clustering of attitudes on a variety of public issues. acts. In ) outset. cognizes III. PROCEDURE ude and of rela- The project conceived in L929 by Gardner Murphy and the attitude present writer aimed flrst of all to present a wide arcay of re, how- problems having to do with these flve major "attitude areas": In the international relations, race relations, economie coirflict, po- I sophis- litical conflict and religion. The attitude areas best covered in rony has the questionnaire are those of race relations, international re- gard the lations, and economic conflict. It was our conviction that very ro regard high specificity would exist among the elements used in our 1) insist- questionnaire, except insofar as the questions clearly dealt ronstitute *ittr the same issues. On the basis of results obtained by .clency in C. W. Hunter (15), for example, in the field of Negro-white e various relations, it was believed that attitudes toward segregation, ;o a wide- toward eating with the Negro and toward lynching would be cceptance independent, and that in general any one speeiflc attitude to- ;ive value ward the Negro would bear no clear relation to the attitudes rpositions on other issues. This, of course, does not mean that we ex- pected to obtair: the rather ideal specificity which would be mstances. character indicated by a zero reliability (by the split-half method) when uency by comparing pro-Negro attitudes on some items with pro-Negro ism, some attitudes on others. It does mean, however, that we expected l and con- fifteen attitudes dealing with nearly fifteen different issues in cal mech- which the Negro is involved, to give exceedingly low split- i early as half reliability, and that in order to obtain anything like a iical char- true "scale" for the measurement of such a thing as pro- or with anti-Negro feeling, one would have to employ dozens or even asted l. w. scores of questions. In the same w&y, the specificity of opin- A11- .y intellec- ions on international and economic issues was assumed to be ,omic, and so great that an enormous number of items would have to be used in order to give a satisfactory reliability for any scale stical con- which could be constructed. It is important to emphasize -generalitY these expectations at the outset, since the results, although in agreement some respects confirming our belief, have in general yielded r, to make remarkably clear-cut group factors, group factors which in rh methods the light of the small number of items used must be regarded r extensive as having, for the student subjects used, a status as clear as n a varietY that enjoyed by group factors of immediate memory, verbal 11 TECHNIQUE FOR THE 12 ability, and the like to r,vhich a number of statisticians have called attention. The method by which the questionnaire was constructed was as follows. Having determined to study intensively the mat- ter of international, inter-racial and economic attitudes, and, to a minor degree, political and religious attitudes, among large numbers of college students at typical American uni- versities, a survey was made of the questionnaires already ad- ministered by other psychologists for these purposes. Among those which proved especially helpful were those of G. B. Neu- (9). mann (23), C. W. Hunter (15) and R. W. Georse In addition, about two hundred newspapers and magazines were . rapidly surveyed during the autumn of 1929, declarations of YXrpinion being culled for consideration, special emphasis being given to the more dogmatic types of opinion frequently found in editorials. A small number of questions were in- cluded from books, addresses and pamphlets, and a number were made up by the experimenters. Wherever it was possible to use questionnaire material which had previously been ex- tL t tensively tried out, and where, in a sense, "norms" were avail- able, we preferred to use the questions exactly as they stood. \ In a few cases, it was necessary to abbreviate and simplify \ the questions in order to make sure that only one issue was t involved and that ambiguity was avoided. In those instances 'in which we made up our own questions, we sought to empha- size simplicity, clarity, and brevity. . Without exception, the questions were presented in such a ' \form as to permit a "judgment of value" rather than a "judg- ' ment of fact." Phrases such as "The United States should," or "We ought to," or "No man should be allowed" constantly reappeared. In a few instances it may seem on first inspec- tion that a question has to do with a question of fact, but closer analysis will reveal the highly arbitrary character of such "facts." Perhaps the least desirable of all the questions used was the following: "Is war at present a biological neces- sity?" Such a question appears to many minds to be categori- cally a factual one; for example, from a neo-Malthusian point of view it may be regarded as capable only of an affirmative answer. The term "necessity," holvever, refers here more to the student's attitudes toward various wants than to any of those types of necessity which are discussed by physicists or