Statistical Studies of Social Attitudes and Public Opinion
Stuart A. Rice
STATISTICAL analysis is only employed where a plurality of individual items or units is involved. It is commonplace to say that the first task of the statistician is to define his units.[1] It is not so often pointed out that units are frequently arbitrary, and that they may depend upon a prior conceptual, analytic, or philosophical process. A given field of investigation may be so formulated, on the one hand, that a plurality of units does not appear. This formulation will preclude the use of statistics, or at least direct attention away from the possibilities of statistical statement. On the other hand, the formulation may be such as to render the statistical mode of attack inevitable and imperative, or at least to call attention to its possibilities and advantages.
Consider an illustration which will anticipate, by a few pages, the later discussion: In some Pennsylvania communities it is believed that if the shavings resulting from the manufacture of a coffin are not swept up and placed therein prior to a burial their presence among the survivors will bring further death or misfortune. If, without further instructions, a teacher were to ask a class of graduate students, "How would you proceed to investigate such a superstition as this? " he would
( 172) probably receive two types of replies: a portion of the class would propose to inquire into the historical circumstances under which the superstition arose and under which it was evolved and diffused; another portion of the class would suggest attempts to explain why the superstition should gain credence in human minds. Neither group would be likely to propose statistical modes of inquiry, because the historical and psychological approaches are suggested when the belief is labeled with the concept superstition. Suppose, however, that the class is asked, "How would you proceed to investigate such a superstitious attitude as this?" Without any restatement of the data and by inserting the single word "attitude," the teacher has changed the connotation of the question and suggested a new type of inquiry. The word "superstition" seems to refer to a discrete psychological entity, one which is either held or not held by individuals. The word "attitude" evokes a concept of something variable, to be examined statistically. The extent to which the two questions actually evoked these two concepts, respectively, would, of course, depend upon the extent to which the suggestions which the terms contain were uncritically accepted in the minds of the students. For instance, one might say "superstition" and the student of independent type might think: "attitude—hence, variable."
The process of conceptual formulation has been presented so far as one which precedes and directs the choice of further means of inquiry. But the sequence may be reversed. The desire to proceed quantitatively, for example, may bring about a conceptual reformulation of the subject matter. This, we have been told by Mitchell, is something which has actually taken place in the recent development of
( 173) the science of economics.[2] A similar trend is discernible in political science.
This discipline, like history, has until recently been largely concerned with legal formulations; and with unique, non-repetitive situations. When James Bryce discussed "modern democracies" he did indeed utilize what Bernard and Kirkpatrick call "the informal statistical method."[3] But it was more usual for political scientists to discuss single states, and to deal with formal aspects of their constitutional organization and activities. One cannot use statistics readily in explaining the constitutional limitations imposed upon the authority of the Federal Government, or in de-scribing the history of diplomatic relations between the United States and Great Britain. But when the general notion of the State becomes less absolute and more functional, when sovereignty comes increasingly to be regarded as pluralistic, when formalism is replaced by realism as a premise in the observation of events, then statistical method begins to have a rôle to play. Such actual changes as these in the developing concepts of political science have resulted in the use of statistical methods. At the same time, they have been a result of the prestige attached to statistics, especially in the sister subject of economics.
Florence, like Catlin,[4] does not even limit the scope of political inquiry to the State. In his recent important book, Florence sets out "to provide for statistical politics the analytic framework, and the
( 174) apparatus of thought" that have already been so largely achieved in economics. Hence, political science "is interested in the acts or behaviour of men toward men, their mutual interrelations and reciprocal contacts; the orders, punishments, votes, verdicts. appointments, dismissals, passing transitively from men to men, and the meetings and discussions between men, that form part of their relations of ruling, manning and sharing of work"(364). So viewed, statistical politics becomes not so much a possibility as a necessity, and the author is able to itemize in considerable detail the statistical inquiries which are required to fill in the frame.
When we come to the subject of attitudes and opinions, then, the first task appears to be an examination of the concepts, and attendant definitions, upon which statistical treatment depends. The process of conceptual formulation has gone on rapidly during the past half decade. It must be admitted that there is still lacking that extent of agreement among investigators which gives assurance that all are talking about the same entities when the same terms are used. (However, this is still quite generally the situation in social science.) Two developments of thought which seem to have laid the basis for the present interest in attitude measurement will be indicated at this point.
The first of these developments has been pointed out by Franklin Fearing in a survey of the subject, as yet unpublished.[5] He notes that present interest in the study of attitudes represents a reaction from the stimulus-response psychology of a few years ago. That is, it represents renewed direction of attention
( 175) writer inward, toward directing motivations in the organism itself, and away from those factors in the external environment that provide the stimulus. By way of definition, Fearing says: "All those factors involved in the readiness or preparedness of the organism to respond are referred to under the terms which we have used, i.e., determining tendencies, mental set and attitude. These include all the processes preceding and determining the motor side of the act, not excepting those processes which antedate the stimulus itself. The latter becomes merely a releasing mechanism."
This might seem to remove investigation entirely from the sociological to the psychological level. It would do so except for the second of the developments to which reference has been made. This appears when the concept of attitude, as just set forth by Fearing, is compared with the concept of instinct, recently so prevalent. In once more directing his attention toward the drives within the individual, the investigator has not returned to the instinct hypothesis. He has taken with him the sociological concept of personality as a resultant of social and cultural experience. Instincts and attitudes would otherwise be identities concealed under different names. Instincts are conceived as inborn; attitudes as composites of inborn drives and experience. Thus, F. H. Allport in a publication of 1924 alludes to "the motor set thus built up by suggestion" (i.e., by experience) which "we may call an attitude." [6] The same distinction is found in a more recent "temporary" definition by another psychologist, G. W. Allport. He refers to attitude as "a disposition to act which is built up by the integration of numerous
( 176) specific responses of a similar type, but which exists as a general neural "set,' and when activated by a specific stimulus results in behavior that is more obviously a function of the disposition than of the activating stimulus. The important thing to note about this definition is that it considers attitudes as broad, generic (not simple and specific) determinants of behavior." [7] Attitudes are "built up by ... responses" not given in the germ plasm at the start of the individual's life.
The consequences for our subject of this distinction between instincts and attitudes are important. The concept of instinct puts an emphasis upon the discrete character of the entity. The concept of attitude emphasizes its variable character. The distinction is not, however, logic-tight. C. Kirkpatrick and J. W. Woodard have called my attention to investigations posited upon the assumption that instincts are variable, and a suggestion of similar import has been made by W. F. Ogburn in his book, Social Change. The most that can be contended is that the concept of instinct does not seem to favor such investigations, while the concept of attitude does favor them. For if we start with the concept of specific instincts, as did William McDougall, our major problem becomes one of identifying these instincts. We are interested in whether or not certain drives are to be found within us, rather than in variations in the intensity of these drives. But if we conceive of attitudes as resultants, in part, of experience, we necessarily have our attention focused upon their variability. Since the experiences of people vary, so must also their attitudes, if the latter grow out of the former. When con-
( 177) -fronted with a given stimulus, which more or less resembles a greater or larger number of other stimuli hitherto responded to, the individual attitudes evoked in a number of persons must be conceived as forming, potentially at least, a continuum along a scale, when evaluated with respect to any quality such as direction or intensity.
Two other theoretical considerations are relevant to the discussion at this point, before taking up the topic of opinion. Some years ago Faris raised the question whether' in discussing instincts, we were dealing with data or hypotheses.[8] It is sometimes suggested that the same question is now timely with respect to attitudes. Are attitudes hypotheses rather than data? The question seems debatable, and perhaps the answer turns upon definition. At the same time, an affirmative answer would not be especially damning. The concept of attitude, hypothesis though it may be, provides a methodological postulate upon which to base interesting and useful classifications of human behavior. Positing the attitude, we seek to find behavior which we may attribute to it. Measurements of the behavior then provide us with indexes of the attitude, which is thereby defined in terms of the behavior. It is in this sense that it seems to the writer to be permissible to say that attitudes can be measured. Moreover, it seems a useful assertion to make, provided the conceptual or hypothetical nature of the terms be not forgotten. It is more practicable to call a man "a conservative" than to describe on each occasion the behavioristic class into which we would put him on the basis of what he says and does.
( 178)
The second theoretical consideration concerns the possibility that attitudes (and opinions) might be studied statistically as they are found at different times or under different circumstances within a single individual. This suggestion is implicit in Thurstone's formulation of the "law of comparative judgment."[9] No instructor, for example, can regrade a set of examination papers with his first marks obscured and award precisely the same grades on both occasions. In a series of decisions affecting the rights and privileges of labor, we might expect an honest and seasoned jurist to be at times more favorable and at times less favorable to labor's cause. The curve expressing the distribution of such variations might, of course, be distorted from the familiar bell-shaped form because of the cumulative influence of habit or of varying trends in the learning process. In this case, perhaps the concept of a time series of attitude expressions, to be measured by a trend line, would be more applicable.[10] However, the present paper is concerned, not with variations of attitude within individuals, but with variations of attitude among individuals. This lightens the task of establishing units that may be statistically enumerated, for while the psychological states with which we are concerned may remain obscure, the human "carriers" of these psycho-logical states are distinct. As Thurstone has pointed out, the unit may consist either of a single average expression of attitude upon a given topic by a single person; or it may consist of any one among a plural
( 179) number of expressions of attitude upon the topic by a single person. That is, there may be exactly as many attitudinal units to enumerate as there are persons; pr there may be more attitudinal units than persons. ' In either case, the process of identifying and counting the units is vastly simplified by their identification with persons.
The argument with respect to attitudes up to this point may now be summarized: We are dealing with psychological entities, "real" or "hypothetical," conceived as variable among individual persons, the essential nature of which if "real" we know little about, but which are brought to our sense perceptions by behavioristic evidences [11] which may be distinguished into units, classified, and enumerated. Behavioristic evidences have not yet been discussed. The topic brings us to a consideration of opinion and its relation to attitude.
In the first noteworthy American effort to measure the distribution of opinions and to indicate their specific relationships to attitude, Allport and Hartman [12] do not develop clearly the general dependence between the two concepts. At the Round Table on Political Statistics of the Second National Conference on the Science of Politics,[13] protracted discussion produced agreement on three points only concerning a definition of opinion: (1) "It need not be the result of a rational process; (2) it need not include an awareness of choice; and (3) it must be sufficiently clear or
( 180) definite to create a disposition to act upon it under favorable circumstances." Further, " On the question when is opinion public, the round table was unable to come to a definite conclusion. "It will be noted that this attempt to define opinion produced a concept strongly resembling that of attitude, as previously defined. In connection with efforts to measure various political phenomena, the writer expressed a preference "to avoid the use of the term opinion and to use instead the word attitude, as indicating a disposition or set toward behavior without reference to the degree of rationality that may be present in connection with it. .. . `Opinion' and `public opinion' seem . . . to connote too much of the rational and conscious elements in the actual motivation." He therefore proceeded to "use the word attitude in a somewhat inclusive sense, without endeavoring to determine in any particular case whether the word opinion might be preferable." [14]
A formal attempt to relate the concepts of opinion and attitude for methodological purposes has been made by Thurstone.[15] "The concept `attitude,' he says, "will be used here to denote the sum total of a man's inclinations and feelings, prejudice or bias, pre-conceived notions, ideas, fears, threats, and convictions about any specified topic. . . . The concept `opinion' will here mean a verbal expression of atti-
( 181) -tude." This distinction has an obvious methodological advantage. Opinions are accessible. Either the opinion itself or the person holding it may be treated as a unit, and classification may then follow, thereby permitting statistical enumeration and analysis. If opinion is regarded as an index of attitude, which Thurstone regards as below the verbal though not necessarily below the reflective level, then attitudes gain a reflected objectivity and precision from the opinions which represent them.
A criticism of Thurstone's concept of the relation-ship between attitude and opinion will be withheld for a little, in order to proceed directly to an examination of the essentials of his method. His procedure is the latest attempt, as it is the most complete, rigorous, and rational, to employ statistical methods in the study of attitudes. It will serve in this discussion as the best example of an important and standard type of inquiry in this field.
Thurstone's most important achievement is his method of constructing a rational scale of values, to which any number of relevant and unambiguous propositions falling in the linear dimension of the scale may be related, each with its own scale value. This has not hitherto been accomplished in attitude measurements. The so-called scales established by predecessors have merely consisted of an arrangement of statements in rank order. The intrusion into the series of new statements, in this latter case, would have extended the scale by a corresponding number of points, and thereby have disrupted all of the so-called values with the corresponding frequencies, previously located upon it. As Thurstone points out, the numbers of endorsements received for the various statements on such a rank-order scale do not constitute
( 182) a frequency distribution in a statistical sense. In Thurstone's scale, on the contrary, propositions may be added or taken away without affecting the scale itself, which is therefore independent of particular statements.
The technique employed in Thurstone's method represents, in a broad sense, an endeavor to apply in the fields of attitude and opinion principles developed by psychophysicists. In essence, the method consists of a determination of the intervals which appear equidistant in the average opinion of several hundred judges. A large number of statements, which are intended to include extreme positions at both ends of a linear continuum of opinion, together with inter-mediate positions, are sorted by each of the judges into eleven classes. These classes appear to each judge to be equally spaced between the two extremes. Each statement will have as many ratings between one and eleven as there are judges. The accumulative proportions of the ratings received by each statement in each of the eleven ordered positions are calculated. These proportions are plotted on rectangular axes, in which the eleven scale positions appear as abscissas and the accumulative proportions as ordinates. The resulting accumulative curve for each proposition, similar to the familiar ogive, is smoothed. The point on the X axis where the curve shows an interpolated accumulative total of fifty percent of the cases is taken as the interpolated scale value of the statement. This scale value is in terms of the eleven equidistant positions into which the base line has been arbitrarily divided. The scale value thus appears as a median valuation by the judges, graphically determined.
The interquartile range, or Q-value, is taken as a measure of ambiguity. "If a statement is very
( 183) ambiguous, the different readers will place it over a wider range on the scale and the Q-value will be correspondingly high." It may then be rejected in the selection of statements to compose the final scale which is to be used in measuring the attitudes of a group of subjects. Thus, a statement is included in the scale only when the judges are in relatively close agreement as to the position which it should occupy with reference to other statements, and with reference to the extremes of attitude.
The objective criterion of ambiguity which is afforded by the Q-value somewhat mitigates, but does not wholly overcome, what seems to the writer to be an inconsistency between two of the assumptions underlying Thurstone's method. This concerns the relationship of opinions and attitudes. When judges are asked to sort statements into classes, representing equidistant positions along a linear scale, they are asked to make judgments concerning attitudes, not to express their own attitudes. The judgment expressed may presumably be called an opinion. For example, among a series of statements it is one's opinion, as a judge, that statement A is a reflection of a more pacifistic attitude than is statement B. The judge's own attitude may be quite militaristic, but he is asked to express opinions concerning the comparative pacifism exhibited by statements at the other end, and at all points, of the scale. The task given him assumes that he is capable of expressing a rational opinion concerning the opinions upon the issue in question, independently of his own attitude toward the latter. All of this is in his capacity as a judge. But suppose his own attitudes are then tested by means of the scale that he has helped to construct. It is now assumed that his opinions are no longer
( 184) independent of his attitudes, but, on the contrary, that the former are indexes or reflections of the latter. In the first case, his opinions must be rational; in the second case, they must not be rational. It should be clearly noted, however, that this criticism depends upon the legitimacy of referring to judgments as opinions.
What is the possibility that the acceptance or rejection by a subject of a statement upon the completed scale may represent a rational judgment concerning the truth or falsity of the statement made? It would seem to exist. If so, the validity of the statement as an index of attitude is destroyed or impaired. If I accept the statement that H2 and O when combined will produce the chemical combination known as water, this does not express an attitude. It is a statement of fact. Might it not be considered a statement of opinion, if there were a degree of uncertainty upon the matter in my mind? If I concur in the proposition that the corn borer is threatening the production of corn in the United States, I might be credited with a factual judgment if I were an agricultural expert. But since I am not, and if I were called upon to support the proposition with evidence, would I not reply that I merely expressed an opinion? My evidence is a bit hazy. My assent to the proposition does not reflect an attitude toward the corn borer, for I may be either a bull or a bear in the corn market. But if I concur in the proposition that "the organized church is an enemy of science and truth," 16 this is taken as evidence of an attitude (having a scale value of 10.7, or a position hostile to the church) in a scale between hostility and favor
( 185) toward the church. Where is the distinction to be drawn between factual-judgment opinions and attitude-representation opinions? The objective criterion of ambiguity does at least eliminate statements upon which the judges cannot agree closely as to a specific relationship between the opinion and an assumed position on the attitude scale.
Another criticism concerns a subtle ambiguity in the use of the terms "favorable" and "unfavorable" in describing the respective halves of the attitude scale on either side of the neutral point. These terms are not used by Thurstone, but have been employed by others who follow his general procedure. It is possible to think of attitudes in terms of the effect they will have upon the person, policy, or situation toward which they are directed. This is not always easy to distinguish from what may be paradoxically called the intent of the motor set within the subject himself. For example, if I am serving as a judge in the construction of a scale designed to measure the attitudes of white people toward persons of color, the effect upon the colored man may serve as my criterion in distributing the statements into the various classes. In the case of propositions A and B, I may reflect as follows : Proposition A, if generally held by white people, would result in benefit to the Negroes. Proposition B, if generally held, would result in less benefit. Hence, proposition A is more favorable to the Negro than proposition B and should be placed at a higher (or lower) position on the scale. The arrangement of propositions, then, is determined by the state of the object rather than by the state of the subject.
To be specific, let us suppose that the judge is a psychiatrist. He is asked to classify statements concerning the Negro into eleven classes; from Class One,
( 186) the least favorable, to Class Eleven, the most favor-able. Suppose he is confronted with the statement: "I regard the Negro as I do myself, and will share with him whatever I may possess at all times." Is this statement favorable or unfavorable to the Negro? The psychiatrist may regard this as definitely unfavorable, because he thinks that the Negro should be encouraged to develop greater independence than such an attitude by whites would permit. Or, suppose the statement to be: "I am impatient of the Negro's dependence upon sympathetic white men. Force him to work and take care of himself." "Ah!" the psychiatrist might rationally say to himself, "that is just what the Negro needs to throw off the psycho-logical heritage of slavery. That is a favorable attitude!" And would he not be right, assuming the correctness of his premises? The attitude would be favorable toward, i.e., result in benefit to, the colored man. But the psychiatrist would not be building a scale of attitudes, as generally understood, and in the process of scale-building the order of the two propositions cited would probably be reversed. By attitudes we intend to refer to, and should employ terms which express, the disposition of the subject rather than the favorable or unfavorable effect upon the object. Perhaps we might say "favorably disposed" and "unfavorably disposed," as indicating a motivation to favor or the reverse, quite apart from any appraisal or calculation of the probable outcome. In his scale of attitude toward the church, Thurstone uses the terms "appreciation of" and "depreciation of." These seem unexceptionable.
It may be added, without going into further particulars, that Thurstone has exhibited great skill and
( 187) care in developing checks and measures of reliability for his calculations at all points.
There is still another conceptual problem involved in the composition of the board of judges. At least three possible types of selection may be considered. In the first, the judges are presumed to be experts upon the issues. A second mode of selection calls for a widely selected sample of persons from the general population. A third calls for a sample from the group or class, the attitudes of whose members are later to be tested. Certain difficulties attend each of these modes of selection.
In studies preceding Thurstone's, the persons who have arranged the order of the propositions composing the so-called scale have been presumed to be experts, as, for example, teaching colleagues of the investigator, supposedly familiar with the subject matter contained in the test. Such judges are accustomed to rational reflection upon such issues, and, in particular, upon the immediate issue. Hence, they will tend to classify attitudes as favorable or unfavorable, rather than as favorably disposed or unfavorably disposed, in the sense of the distinction drawn above. The very possession of expertness, then, may be viewed as tending to render the judge incapable of arranging propositions in such a manner that they will seem to non-experts to be in order and to have linear differences equal.
An analogous difficulty applies to the selection of judges who will constitute a sample of the general population. It may be contended that not only the "equal-seeming interval" but the order of propositions itself may differ considerably among different groups, on the average, as a result of differences in cultural background. An opinion characteristically
( 188) regarded as favorable, or favorably disposed, to the Negro in the South might be characteristically regarded as unfavorable, or unfavorably disposed, in the North, and vice versa.[17]
The third of the three choices proposes the selection for judges of a sample from the same class or group or segment of the population as that to which the scale is to be applied in the subsequent test. It shares another difficulty with the second choice, just examined: Will persons in any sample made up of non-experts have sufficient understanding and command of abstractions, and particularly of language, to verbalize their attitudes? That is, will the average judge in the sample be intellectually capable of making the discriminations called for by his instructions and detaching his own attitudinal bias from the discriminal process? Is the average person competent to perform this dual and rather dexterous task? Perhaps the answer is that these questions apply with much the same force to the subjects to whom the scale is applied. Perhaps the attitude scale, like other scales employed in science, is valid only within the middle ranges of its phenomena—in other words, among persons whose intellectuality and knowledge are neither too great nor too small.
There is still another disappointment in store if the third mode of obtaining judges is employed. One of the major objectives of the method is to compare the attitudes of different groups. If each group requires a measuring scale built up for that group alone, direct comparisons between groups as to average tendencies become impossible, although comparative variability might still be determined.
( 189)
Unsatisfactory as is this third among the choices, because of the limitations that it seems to place upon the use of the scales, it seems to involve less contradiction than do the first and the second. If so, we must be contented to recognize that a scale devised with the aid of judges in Boston, for example, when applied in Atlanta, would interpret attitudes solely in terms of, or relative to, a Bostonian culture pattern.
It may be comforting to recall that similar considerations have often been overlooked in intelligence measurements. Even the familiar scales of physical measurement are regarded by physicists as relative, and as inapplicable when the physical constants change appreciably. Thurstone has at least made a be-ginning, and it seems to the writer a first beginning, in the exact and rational application of psychophysical theory to the construction of an opinion scale, with resulting occasion for the use of statistical methodology.
Statements of opinion, however, are to be regarded as but one among various forms of expression of. attitude. Non-verbal, or more accurately, non-propositional expressions may likewise be susceptible of classification, and hence permit of counting and statistical analysis. Again, we may classify attitude studies according to the degree of control which the investigator is able to exercise. We thus have at least four types of actual or possible attitude studies which might receive attention in the present survey. Thurstone's studies are of the controlled verbal or propositional type. The studies of "social distance" inaugurated by E. S. Bogardus,[18] and the study by Donald Young [19] of the effects of classroom instruc-
( 190) -tion in changing student attitudes with respect to race differences, might possibly be cited here as illustrations of controlled but non-propositional studies. Others, more definitely of this type, may easily be conceived. The late lamented silent drama, for instance, presented many situations in which the attitudes of an audience were tested and might have be engauged by such indexes as the sound volume of applause or the ratio of disgusted patrons walking out on the show.
It is in the realm of non-controlled data, whether semi-propositional or otherwise, that the practical need of further and continued attitude studies seems to the writer to be most evident. The difficulties of building scales similar to Thurstone's, and of applying them to the measurement of the attitudes of social groups, become increasingly difficult once we leave the classroom, the discussion club and the other small, comparatively infrequent and highly selected groups that enjoy having experiments tried upon them. Such groups already have developed ways of making their attitudes articulate. It is the more numerous work-a-day groupings of society, which are inaccessible to his controlled measurements, about whose attitudes the social scientist is in the most need of information. Students may be required, good natured academicians may be cajoled, and sundry needy persons may be paid to sort cards containing propositions into eleven piles. But it is difficult to imagine securing comparable judgments, or satisfactory measurements in the final application, from bricklayers, business men, Italian-Americans, nuns, stevedores, or seamstresses. And, unless the scale itself is based upon equal-
( 191) seeming differences to a random sample of the group which is to be measured, its validity—the degree to which it measures that which it purports to measure—becomes open to question.
There is need, therefore, to examine more carefully the extent to which data established for other purposes, but which to some extent reflect attitudes, may be used as indexes for measurements of the latter. In illustration, such data may be cited as political campaign appeals; election registrations and voting returns; newspaper editorials, news content, advertisements and circulation; purchases of commodities which are dependent upon certain activities, interests, or social habits, such as bibles, bathing suits, hymn books, contraceptives, football tickets, bath tubs, listerine, etc., etc.[20]
Further progress in the statistical attack upon atti-
( 192) -tudes and opinions depends upon continued development in several directions: (a) upon laboratory studies of individual psychology which will aid in further clarification of concepts; (b) upon attempts at further empirical verification of the Thurstone scales; and (c) upon ingenious statistical analysis of innumerable "uncontrolled" by-products of human activity, which throw light in specific fashion upon the motivations in our minds.
[Editors' note:
Click here for response from Thurstone.]
