Chapter 20: Supplementary Chapter A: The Technique of Attitude Measurement
Daniel Katz and Floyd H. Allport
THE foregoing chapters have given an account of an attempt to describe the
attitudes of students in a large American university. We have, in this work,
made an effort to apply to the field of attitudes and opinions the methods of
quantitative description familiar in science. It may be assumed that the reader
who has patiently followed thus far is probably motivated by an interest
extending beyond the significance of the data actually presented. Such an
interest of the reader may be either theoretical or practical. It may lie in the
direction of the technical problem of attitude measurement as such, or of a
wider knowledge of the behavior of college students. We are adding, accordingly,
two supplementary chapters. The present one deals with current methods in the
quantitative study of attitudes and exhibits the relation of our own techniques
to those employed by other investigators. The final chapter will summarize the
results of attitude studies in various colleges and will compare their findings
with those of the present investigation.
1. SUGGESTIONS REGARDING THE DEFINITION OF ATTITUDE
We may turn first to the definition of the word 'attitude'. Symonds has called attention to seven different meanings of the term in psychological literature. (58) It may refer (a) to the organic drives more familiarly known as purposes or motives, (b) to muscular set or adjustment, (c) to generalized conduct, (d) to neural set, (e) to the emotional concomitant of action, (f) to the feeling concomitant of action, and (g) to certain verbal responses indicating liking or disliking, acceptance
(354) or rejection. Because of this great variation in usage and be-cause "attitude" seems to him to add nothing to our conceptions of "habit" or "skill", this writer suggests abandoning the term altogether.
But Symonds, like those who have confused the meanings of the term, has overlooked an essential point of method. The above mentioned differences in definition are due in large part not to an intrinsic ambiguity in the material studied, but to differences in the aims of the investigators. On the one hand the experimenter may regard attitudes in terms of psychological and physiological processes, such as emotion or neuromuscular set. He will be interested in attitudes as such, in much the same way as he is interested in thinking or learning. On the other hand, he may be concerned only with the social or intellectual content of his material. He will then define attitudes in terms of certain specific verbal responses,—reactions, for example, which exhibit radicalism, fair-mindedness, or respect for law. He will then be interested not in attitudes as such, but in the particular attitudes possessed or revealed by an individual. The difficulty in attitude study therefore lies partly in the fact that investigators have shifted their approach from one conception to the other without realizing what they were doing. This is a serious error. A physiological process studied by the methods of natural science should not be confused with social valuation or meaning. One cannot expect the same possibilities of measurement and quantitative treatment in the latter field that one has a right to expect in the former. An investigator concerned with respect for law or the belief in a high tariff is working in the field of meanings, valuations, and telic concepts. He must employ different methods and different concepts of denotation, interpretation, and prediction than he uses when he investigates the neuro-muscular or visceral processes involved in the holding of the attitudes concerned.
While no sharp line of demarcation can be drawn between attitudes and personality traits, the latter should in general be distinguished as denoting characteristic forms of behavior, whereas attitudes are more frequently regarded as sets for certain kinds of verbal response expressing value. "Trait" denotes a type of behavior which the individual exhibits in many
(355) differing situations; "attitude" signifies the set to respond in a particular type of situation. This set may or may not be well integrated with the remainder of the individual's personality.
2. INFORMAL QUESTIONNAIRES; 'YES' AND 'NO' PROCEDURES
Perhaps the simplest way to gather attitudes is to ask people directly what their attitudes are. This has been the common procedure in questionnaire practice. A direct, printed question is presented and the subjects respond with a free, written answer. E. S. Conklin has used this procedure in his study of superstitious beliefs and practices among college students. (18) The value of this simple method depends almost entirely upon conditions which are not a part of the method itself. That it can be used successfully when such conditions are controlled is suggested by Conklin's results. He found, for example, that the superstitions of women relate primarily to the home and to social affairs, while those of men relate more generally to sports and business, a fact which seems to accord with the known tendency toward a more highly personalized type of behavior among women than among men. Frequently the direct question and answer is used to supplement other devices in questionnaire forms. This was the case in the Yale Student Survey which was carried out under the direction of A. B. Crawford and provided data of value to the personnel department in that university. (21)
Obviously the method of the simple question and free answer has many disadvantages. The data gathered in this manner are cumbersome to handle. The stimulus situation is not standardized for all subjects. The answers given represent a complex and unanalyzed group of influences and shades of meaning. We have scarcely the slightest approach to measurement here ; and it is therefore difficult to use the data for comparative purposes.
To facilitate the handling of data there has been devised the familiar procedure of using questions calling merely for an affirmative or a negative reply. This is the practical method of politics, where people must commit themselves by voting on an acceptance or rejection of candidates or of issues. Willoughby has reported such a sampling of student opinion at Stanford.(79)
Fenton, in studying honesty in examinations, supplemented his objective study with questions, to be answered "yes" or "no", on attitudes toward cribbing. (24)
Another use of the 'yes' or 'no' question is the search for more generalized attitudes by the summation of affirmations and negations of a number of questions. Investigators have chosen specific questions for summation which were judged by them, or by raters, to be related to one attitude variable. Symonds, in a study of "liberal-mindedness," selected over one hundred questions by having these questions answered from a liberal viewpoint by himself and four helpers. (59) Questions upon whose answers at least four of the judges agreed were retained for the final questionnaire. The students' liberal-mindedness was thus regarded as a function of his acceptance or rejection of these answers. In a suggestive study of radical-ism and conservatism, H. T. Moore used a similar technique. (43) His questionnaire contained twenty 'yes-no' items on the basis of answers to which Dartmouth and Yale students were designated as radical or conservative.
Attempts to add affirmations or negations and to deduce from such a summation a generalized attitude are open to objection. In the first place this procedure generally leaves open the question of how representative and comprehensive are the specific questions with respect to the attitude variable they seek to measure. Is fair-mindedness really adequately measured by the questions included upon a particular test? Secondly, we do not know how diagnostic the different questions are. Some of them may reflect fair degrees of the attitude in question, while others may be wholly irrelevant. Where this method is used there should be validation of the significance of the questionnaire items by other criteria and an appropriate weighting of the specific questions. Through her failure to secure the same results as Moore in his study of radicalism, Washburn demonstrated the part which neglect of such factors as these may play in the conclusions reached. (73)
E. G. Lockhart has made a further development of the procedure of "yes" and "no" answers in his study of children's attitudes toward law. (35) He described fifty-one specific situations in which particular laws were broken, and after each
(357) situation asked three 'yes-no' questions. A score of 1 was given for each situation in which all three questions were answered correctly, the criterion of correctness being the replies of fifty Iowa lawyers. Only those items were included in the test upon which eighty per cent of the lawyers had shown agreement. As a further check of the validity of this standard, the test was given to a group of graduate students. Although Lockhart's study is a more thorough attempt to elicit attitudes for comparative purposes than many earlier researches, it can be criticized for arbitrarily treating the items as of equal value in the score. For while each situation described included the violation of a specific law, the laws varied in range from local ordinances to state statutes. The presenting, however, of specific instances of law violation as the stimulus setting is an effective device, since it helps make constant the situation to which all subjects respond.
3. MULTIPLE-CHOICE METHODS: RATING AND RANKING OF ATTITUDES
Another common technique in attitude studies is the multiple-choice questionnaire, a form in which the subject checks one or more of a list of possible answers. These answers may be listed following a direct question, or following a description of a situation, as alternative responses which the subject considers suitable, or characteristic of himself, in that situation. The most representative use of this procedure is found in tests of personality traits rather than in studies of attitude. The ascendance-submission test devised by G. W. and F. H. Allport is an example of such usage. (3) Wherever the multiple-choice procedure is used, and a number of items each affording selective responses are included, we meet again the problem of the relative value of the different items in the final summation which constitutes the score for the entire test. The solution lies in careful validation and standardization on each item. In the All-port study to which we have just referred such standardization was accomplished by taking the median rating by associates, on the trait of ascendance-submission, of all the individuals who chose a certain answer to the item in question. Anderson and Dvorak have used the multiple-choice method in an attempt to
(358) discover the standards of conduct of college students as compared with the standards of their elders. (5) Fifteen behavior situations were described, under each of which four possible answers were listed, one dealing with the situation according to a standard of right and wrong, another according to a standard of prudence, a third according to public opinion, and a fourth according to an ęsthetic standard. G. B. Watson has employed the multiple-choice technique effectively in his test of fair-mindedness. (74, 75) In form C of this test, for example, subjects were presented a list of conclusions which might follow from a given statement of fact. Endorsement of certain conclusions were regarded as indicative of prejudice. Crawford's study of student opinion at Yale employed a simple form of the multiple-choice procedure. (21) The reader will be aware also of the large part played by this method in the Syracuse Reaction Study.
We have used the term multiple-choice to represent the listing of a number of possible answers for subjects to check. These answers do not necessarily imply a scale or continuum. Rating and scale devices are a special type of multiple-choice procedure which differ from the simple type in that they pre-sent various degrees of a single attitude variable instead of a series of unrelated opinions about a given question. The single multiple-choice questionnaire, like the multiple-choice scale, is obviously superior to the single 'yes-no' questions; but the type first mentioned gives no basis for an intelligent comparison of the answers checked. Many investigators have therefore used some form of rating or scaling technique to present successive gradations of attitude.
In a comprehensive survey of opinion on war, Porter asked students to rate themselves on their degree of agreement with one hundred and fifty statements. (45) The subjects were limited to the following five categories regarding their belief in the correctness of the statement : certainly right, probably right, doubtful, probably wrong, certainly wrong. To establish a criterion of the validity of his study Porter had some of his subjects rated by their acquaintances on a rating scale for their degree of militarism or pacificism. H. A. Sturges had previously used much the same procedure in studying the degree of
(359) change in the attitudes of college students. (57) E. S. Jones had his students mark attitudes on social, religious, political, and economic issues with numerical indices of their belief in the truth of the statements given. (31) T. H. Howells has carried the same technique further by using point scales on which students rated their own attitudes on about one hundred and sixty items, the crucial ones centering about religious orthodoxy. (29) On each question the student was asked to express his degree of belief in comparison with the position he considered to be held by the average person. The rating scale technique has been used by Lund in studying the psychology of belief. (36) Students were given thirty propositions on matters of religion, ethics, politics, and science, and were asked to mark each one on a scale ranging from +10 to -10 according to the strength of their belief or disbelief. A rating was also obtained on the degree of certainty of their convictions, and a third on the desirability of the proposition under consideration. In this experiment, as in others, the subject is asked for an introspective report of the intensity of his feeling. Dunlap and Snyder, in their study of the moral evaluations of college students, used a rating technique with a somewhat more objective emphasis. (53) One hundred statements of actions were presented. The subjects characterized them by a score of -10 for the worst action, and a score of +10 for the best action, with intermediate acts rated accordingly. The instructions were, first, to find the worst act, the completely neutral act, and the best act, so that these might be used as standards for the rating of the others. The average value for each action was computed to determine its place upon a scale of moral judgment. Although the authors regard their work as merely preliminary because of the inadequate number and range of their subjects, this experiment is significant as one of the first attempts at the construction of an attitude scale.
The application of the rating technique to attitude research can be criticized on the following ground. While it does in a measure reflect gradations of attitude, it fails completely to standardize and objectify the stimulus situation. An individual in indicating his own position, for example, upon a radical-conservative rating scale, may be a poor judge of his own
(360) characteristics; but more than this he may not be using the same criterion of judgment that another subject might use with regard to the meaning of the different positions on the scale. If every individual is allowed to determine his own measuring stick each set of measurements will belong in a category by itself.
A more objective approach to the attitude-rating procedure has been devised by E. S. Bogardus. In his well known studies of "social distance" this investigator asked subjects to indicate their attitudes toward races and nationalities on a scale consisting of seven degrees of social intimacy. (10, 11) By listing seven specific degrees of social distance this method achieves a certain standardization of the situation for all subjects, so that one can compare the attitudes of two or more subjects without the assumptions implicit in the rating scale where subjects merely give a symbol representing the degree of their feeling. On the other hand, Bogardus has made the unwarranted assumption that by calculating numerically the degree of attitude as determined by the steps of his scale he can give a true measure of that attitude. He has, in reality, no genuine measuring scale, but only an a priori device which cannot guarantee the equality or determine the possible inequality of its successive steps. Such a device is legitimate for evoking attitudes and even for a comparative study; but it is legitimate only when the groups to be compared are considered step for step, with no at-tempt to give average scores of social distance for the respective groups.
The ranking as well as the rating method has been used to determine attitudes. Occupational preferences of college men were found by Anderson by having students rank twenty-four occupations. (6) Clark listed social offenses of varying severity and asked his subjects to place them in rank order.(17) Brogan had students prepare a list of bad practices with which they were familiar and arrange these practices in rank order according to their seriousness. (12) These studies are not attempts to devise scales by the method of rank order. The technique is used merely to reveal the attitudes of the students observed. As we have already noted, however, the observations of such attitudes can readily be used for constructing a scale in the field concerned.
4. THE LOGICAL, A PRIORI, ATTITUDE SCALE
One of the first attempts at an objective scale for the measurement of attitude was the Allport-Hartman study. (1, 2) The authors describe their experiment as follows : "Seven concrete issues of current interest were chosen, dealing, respectively, with the League of Nations, the qualifications of President Coolidge, the distribution of wealth, the legislative control of the Supreme Court, prohibition, the Ku Klux Klan, and graft in politics. Sixty students, upperclassmen, were asked to write their personal views on the various phases of these questions. The resulting opinions on each issue were then care-fully sifted and the distinct and relevant views were assembled. Keeping the issues separate, these views were printed on slips of paper and arranged independently by six judges, teachers of political science and psychologists, in order of their logical position in a scale ranging from one extreme on the issue in question to the opposite extreme. The average rank assigned to each statement was taken as its final rank in the completed scale."
By this method Allport and Hartman devised scales for the measurement of opinion on seven issues. These scales were then given to students for checking in order to ascertain the distribution of their opinions on the attitude variables concerned. Besides their interest in a device for measuring opinion, the authors were concerned with consistently atypical individuals and with the factors in their personalities which might be related to their atypicality. The scales developed in this study really represent mere rank orders. There is no rational unit of measurement and hence no way of determining the equality or inequality of the intervals between their steps. The statements representing the successive steps, moreover, are not always phrased so as to represent varying degrees of a single attitude variable. In regard to logical construction, therefore, the scales of the Syracuse Reaction Study are a considerable improvement over this earlier attempt.
About the time of the Allport-Hartman study, or perhaps earlier, Willey and
Rice reported on changes in students' opinions regarding evolution following
upon an anti-evolution lecture delivered by William Jennings Bryan. (78) Five
statements of logical degrees of acceptance or rejection of the evolutionary
doctrine were used as a scale. G. W. Allport studied the political and economic
opinions of Dartmouth students one month prior to the presidential election of
1928, using four-step scales similar in form to the scales of the Reaction
Study. (4) George Vetter, in an unusually careful and comprehensive study of
radicalism, conservatism, and reactionism, has made an interesting use of the
logical, rank order scale. (72) He developed five-step scales on thirty-six
social and political issues, selecting one attitude variable only for each
scale, and making his scale steps express different degrees of that attitude
from extreme radicalism to extreme reactionism. To make sure that he was not
gratuitously assuming these qualities in the positions of his scale, Vetter
mixed together the statements from all of his scales and placed them in random
order in one list. Eighty-four judges went through this list and characterized
each statement as either reactionary, conservative, neutral, liberal, or
radical. The modal characterization of each statements was used as its scale
value in interpreting the returns later to be obtained from the use of the same
items in scale form. In comparing groups of students as to the incidence of
degrees of radicalism and conservatism among them, Vetter was careful to make
his comparisons separately for each step position in a manner similar to that
which we have employed in the Reaction Study, thereby avoiding the assumption
that the intervals between the steps were equal.
5.THURSTONE'S PSYCHOPHYSICAL METHODS
The investigator who has made the most significant contribution to the measurement of attitudes is L. L. Thurstone of the University of Chicago. Thurstone has brought to the service of this problem the traditional methods of psychophysics. His contention is that the logic of measurement in psychology is essentially the same no matter what function one attempts to measure. To Cattell Thurstone gives the credit of being the
(363) first to extend psychophysical techniques (which were developed primarily for measuring discriminatory functions for simple sensory stimuli) to other fields. (70) Thurstone, however, was the first to see the fruitful possibilities of these methods for the measurement of attitudes.
A mere description of three or four, or a dozen, categories of attitudes, and their incidence in a given group, is, thinks Thurstone, of limited value. We need also to know the quantitative relationship of these categories, that is, how much more of an attitude is represented by one category than by another. We frequently would like to know the factors in a given population which are related to a certain attitude. To ascertain this it is convenient to take two groups which vary in respect to some known factor, such for example as sex, and compare them on the attitude variable in question. For exact comparison it is necessary to know how the attitude variable is distributed in the two groups, what central tendencies exist, and how extensively opinion in the groups at large varies from these central tendencies. It is also of value to be able to give for two groups which are to be compared as wholes, single indices (which reflect their positions more clearly than a step-by-step comparison can reflect them) upon the attitude variable in question. All these problems require for their solution a linear continuum or scale by which opinion can be truly measured. Until we can build a scale whose steps shall be known to be equal, our calculations of the median or mean attitudes of our group and the measuring of dispersion from these central tendencies will be devoid of significance.
The following procedure is illustrative of Thurstone's method. A single, specific attitude variable, for example, the advisability of military preparedness, was selected; and the strength of various statements of opinion for and against it was determined through the judgments of a large number of subjects. These subjects do not indicate their own views, but merely act as judges of the strength of the various statements presented to them as indicating pro- or anti-military-preparedness. Thurstone has found it advisable to use between one hundred and two hundred such statements of opinion in the construction of a scale. The ideal method of obtaining judgments
(364) from the subjects would be that of paired comparison. This would mean that each statement of opinion would be presented to every subject in paired combination, serially, with every other statement, the subject judging which of the two statements represents, for example, the strongest leaning toward militarism. Every subject thus makes as many judgments as there are possible paired combinations of statements. This method is impractical because of time consumed and the fact that the number of judgments required of the subject is so great as to produce fatigue and render the discriminations unreliable. Thurstone has used the method of paired comparisons in constructing a scale for nationality preferences, where the number of judgments was considerably reduced because only twenty-one nationalities were employed. (66) Having obtained a large number of judgments of preference for each nationality as compared with every other nationality, he constructed his scale values on the basis of his "law of comparative judgment". (33) This law enables one to compute the scale-interval separations of statements through the experimentally observed proportions of judgments that one of the statements is stronger than the other. This means, in simpler language, that if a large preponderance of judgments declare statement a stronger than statement b, these two statements are to be spaced rather far apart in the scale. If, on the other hand, barely more than half of the judgments make statement a stronger than statement b, these statements will be placed relatively close together upon our scale.
It is also necessary to take into account the degree of confusion of each of these statements with the other statements in the scale. Thurstone calls the activity by which we compare two stimuli and judge one to be greater or less than the other the discriminal process. The discriminal process for a given stimulus is not fixed. It fluctuates between successive judgments of one subject or among single judgments of many subjects. In other words, stimulus a may be judged greater than b in only 60 per cent of the judgments obtained; but it may be judged greater than f in 90 per cent of the judgments. This fluctuation in the discriminal process for any stimulus will give us a frequency distribution for which we can find a mode and a standard deviation. The standard deviation of such a distribu-
(365) -tion Thurstone calls the discriminal dispersion, and it is this standard error which he takes as a unit of measurement when some form of the constant method is employed. (62)
In another experiment Thurstone used the rank order method to discover the real values of the steps in the prohibition scale of the Allport-Hartman study. (64) Two hundred subjects were asked to place the thirteen statements of this scale in rank order. This procedure is shorter than the method of paired comparisons, and from it can be obtained the data necessary to compute scale values by the law of comparative judgment. Where as many as one hundred statements are used, as in the construction of most of Thurstone's scales, this method, like the method of paired comparisons, is impracticable.
A method which Thurstone finally found very useful is that of equally appearing intervals. (68, 70) Subjects are asked in this method to sort the statements of opinions into a given number of piles. The first pile is used for opinions which seem to the subject to reflect the maximum support or acceptance of the matter to which the attitude variable refers. The middle pile is used for statements which seem entirely neutral. The last pile contains opinions representing the extreme of attitude ad-verse to the matter in question. The intervening piles are employed as equally spaced locations for opinions lying between these termini. For each statement there is thus obtained the number of the pile or step into which it is thrown by each subject. The median step value of all these placements for each statement is computed, and is used as the step value of that particular statement upon the scale. In the final form of the scale only those statements are retained which did not show great variation in the piles to which various judges assigned them. Statements that tend to be widely distributed upon the scale can be eliminated as being ambiguous in their meaning. Thurstone and Chave have used the method of equally appearing intervals to construct a scale of attitudes toward the church (70) ; and a whole series of scales upon social and political questions are now in progress in the Chicago laboratory.
One difficulty in the method of equally appearing intervals is the end effect that inevitably results from the fact that there
(366) is always a first pile and a last pile. The fact that there are no more places to be occupied beyond these end piles tends to skew the distribution of the placement of statements which belong at the end regions of the scale. To remedy this defect, Thurstone has changed the method of equally appearing intervals to a procedure which he terms the method of successive intervals. The sorting procedure here is the same as in the former method, but the calculation of scale values is different. The method of successive intervals corrects for the terminal distortions by so spacing the piles that the statements at the extreme of the scale will still give normal distributions in their discriminations. Thurstone has made this correction because he found that. while the rank order method and the method of paired comparisons resulted in similar scale values for a given set of statements, the method of equally appearing intervals did not agree with the other methods in the case of extreme statements. For many practical purposes, however, the difference is not sufficiently great to invalidate a scale based upon the method of equally appearing intervals.
Another of Thurstone's procedures for obtaining scale values is based upon the endorsement or rejection of opinions rather than their discrimination. (69) This process, known as the method of similar reactions, "assumes that if two attributes tend to coexist in the same individual they are regarded as functionally similar, while if they are more or less mutually exclusive to that they tend not to coexist in the same individual then they are functionally dissimilar. The degree of similarity is measured in terms of a . . . . coefficient which enables us to allocate the attributes along a single continuum, and to measure the degree of similarity by scale separations on this continuum or scale. For the derivation of this coefficient see Thurstone's article on "Theory of Attitude Measurement." (69)
In most of his experiments Thurstone has been careful to devise some check upon the internal consistency of his data. When he employs his law of comparative judgment the check consists of comparing the calculated proportions of judgments
(367) with the experimentally observed proportions. Criteria of internal consistency are of fundamental importance, since the first step in experimentation is proof that the data obtained are a function of some actual psychological process and not an artifact. Thurstone's scales also sample consistently the same function, as is shown by their high reliability coefficients.
An advantage in scales constructed by psychophysical methods over a priori, logical scales is the reducing of the artificiality of the statements which subjects are asked to accept or reject. In an a priori scale the statements often represent intellectually deliberated positions, whereas Thurstone's scales contain the simple, direct expressions of opinion that are heard daily upon all current issues. It is impossible to use such expressions in other than an experimentally derived scale, for in the absence of psychophysical procedure they are impossible to classify with any certainty. The psychophysical methods also obviate the problem of emotional stereotypes or phrases which, through some common bias, receive an overwhelming acceptance or rejection. Such phrases are likely to creep into an a priori scale and there distort the true issue upon which we are seeking to elicit the subjects' responses.
An interesting question in adaptation of psychophysical methods to attitude measurement may be briefly mentioned. While such methods may work very well for simple sensory discriminations, when we are judging values of a social sort we may load the experiment with a constant error if we use as judges individuals from a given cultural setting whose discrimination of attitudes may be a function of their particular social background. If, for example, we take our judges on a radical-conservative issue from a conservative section of society, we may find in their judgments a nice discrimination of degrees of conservatism but a poor discrimination of degrees of radicalism. To the extreme conservative all radicals, or even radicals and liberals together, may look alike. Thurstone and his students have been investigating this problem, and they report that they have found this factor to be of little practical consequence. (50) It is perhaps too early, however, to pass final judgment upon this question.
The greatest disadvantage of Thurstone's methods is a prac-
(368) -tical one, namely, the amount of time and labor involved. This criticism, however, has been somewhat exaggerated. The techniques now in use have reduced the labor more than one would imagine from reading a review such, for example, as that of Bain. (8) They still do not permit, however, the selection of a problem today, the construction of a scale tomorrow, and the publication of the results the day following.
6. METHODOLOGICAL PROBLEMS IN ATTITUDE MEASUREMENT
Thurstone considers that when he measures attitudes he is measuring the degree of feeling or affect toward certain symbols. This conception has two advantages. It relieves him of the responsibility of predicting overt behavior, and it gives him justification for thinking in terms of an actual continuum or scale. Attitudes as the content of social situations or proposals are discrete qualities, incapable of being put, in other than logical or teleological fashion, upon a scale. Attitude as the degree of feeling for or against a symbolized situation permits of quantitative gradation. But, as Thurstone would doubtless agree, we can have a logical as well as an affective or psychophysical continuum. Statements can be worded a priori in such a fashion as to fall on a logical scale with reference, for example, to gradations in meaning upon a given variable. It is true that, in its purely a priori form, we do not know whether there are equal intervals between two adjoining statements of such a scale. We can, however, be reasonably certain that the steps are in the correct order. For some purposes this type of crude logical continuum, which is readily constructed, has distinct advantages. With somewhat more labor it can be developed into a true scale by the use of Thurstone's psychophysical technique, while still keeping it as a logical rather than an affective continuum. There arise occasions upon which we wish to know not only the degree or scale value of an individual's attitude, but its nature or content as a practical judgment or mode of adjustment which the subject would be likely to adopt in a par-
(369) -ticular situation. For this purpose a scale based upon a logical continuum would be useful. Let us take, for example, the differing conceptions of the deity listed in Item 50 of the Syracuse University Reaction Study. These statements are among the most interesting and humanly significant notions of the deity we have been able to find; and we are naturally concerned with the individual's reaction to them in and for themselves, regardless of what such a reaction may signify for a rational scale ranging from pro-deism to anti-deism. This does not mean that we are not also interested in the affective reaction toward the deistic symbol which Thurstone's scales so clearly exhibit. Both sides of the picture are important and should be studied.
We may say that our a priori scale, if scale it can be called, is not a continuum reflecting a single underlying attitude variable, but a series of discrete attitudes logically related to one another. These attitudes are not projected onto a psychophysical continuum, but only related telically in that they are successive acts of practical adjustment that an individual would be likely to perform if headed in his overt behavior or thinking toward one of the termini of the scale. In Thurstone's scales the statements are expressions which reflect the underlying attitude; in our scales (that is, in the scales of the Syracuse Re-action Study) they represent the attitudes themselves.
If teleologically viewed, social and cultural behavior do not run in an unbroken graded continuum. The acts which one is "set" to perform toward a certain objective may become more vigorous and effective as one proceeds toward the limit of things which an individual can do in that direction. But these acts and attitudes are, nevertheless, discrete. Take, for example, the objective of atoning for an injury done to the reputation of another person. There are a number of acts which can be done toward this end; and these acts can, with sufficient care, be logically arranged in the order of their increasing effectiveness as means of making amends. One might, for example, call the aggrieved person by telephone and apologize, one might make a personal call expressing remorse, one might make an abject public apology, or one might offer half of a considerable fortune as a restitution for the wrong done. The acts in such a series are discrete. They are qualitatively differing behavior pat-
(370) -terns, not measurable gradations either of a muscular or glandular process or of an introspective degree of feeling. It may be, moreover, that there are no practicable steps of attitude which lie between them; and we do not know offhand whether they lie at equal intervals along the scale upon which they are placed. But it would be possible, by the use of Thurstone's technique, to arrange such acts with reference to a continuum or variable based upon their relative objective effectiveness for accomplishing the end of the maximum possible atonement. From such a standpoint we are not interested in a series of uniform gradations of apologetic feeling, but in these unique types of action (described above) which the individual has learned, as differing cultural and linguistic habits, representing successive stages of apologetic and restitutional behavior. Again, it should be made clear we are not discounting the importance of a scale based upon degree of affect. For many purposes it would be more important to know how apologetic an individual felt than how apologetically he acted. Both phases have their value for differing purposes in psychology and social science. We should probably gain valuable insight, moreover, from a comparative study through both kinds of continua employed upon the same subjects.
The writers therefore suggest that Thurstone's methods might be adapted here,
with statements representing not opinions to be endorsed, but admissions of
specific acts or tendencies toward action; and that the scale be constructed
using the same technique, but with a criterion not of degree of affect, but of
the comparative value of the statements as means toward accomplishing the
objective which the teleological variable implies. The completed scale, in other
words, might be a means of measuring not affect, but effect. For example, in
arranging the statements on means of handling the prohibition question into
piles, the judges might bear in mind not the standard of degrees of liking,
indifference, or disliking for prohibition, but the standard of what an
individual who believed, say, in complete prohibition and could not quite
accomplish it would try to do next. Should he be unable to do that, the next
pile would contain his second choices from the alternatives given, and so on
until he was forced into the final position of accepting complete lack of
regulation of the liquor traffic. A scale of roughly equal telelogical steps
might thus be constructed.
7.THE ADMINISTRATION OF ATTITUDE SURVEYS
Before leaving the subject of technique in attitude measurement there remain for consideration a few general procedures in administering tests of attitudes. One of the first problems concerns the anonymity of the individuals used as subjects. Some investigators are careful to assure the subject, as we have done in the present study, that there is no way of identifying his questionnaire. Others have felt that students will give more serious co-operation if asked to sign their names, since doing so involves a greater feeling of personal responsibility. To a large extent the answer to this question probably depends upon the nature of the attitude that is being studied. Individuals seem to have two sets of attitudes on many questions, namely, the opinion which they are willing to publish or make known (their public attitude), and the opinion that which they will express only to their intimate friends (their private attitude). In studies where individuals know that their questionnaires will be identified they are probably more likely to respond with their public attitudes. Where the questionnaire form is anonymous we may expect to secure a closer approximation of their private attitudes. Both sets of attitudes are important and are necessary for a complete picture. The only requirement is that we do not confuse them through our technique or our interpretation of the results.
To insure the serious co-operation of the subjects in an attitude questionnaire is often a difficult task, especially in these days when the questionnaire is being overworked. Where little time and effort are required most persons do not object to being drafted. When the study is to be more thoroughgoing, however, some special incentive is desirable. It is better to give this incentive in the form of an interest in the study rather than as mere compulsion. Crawford met this problem by making his Yale survey largely a voluntary student enterprise. (21) In the Syracuse Reaction Study care was taken to arouse the interest of students both by their initial participation in planning it and through publicity in their college newspaper.
The majority of investigators have administered their test forms in the classroom under the supervision of an instructor. This procedure probably tends to minimize listless or facetious answers, as well as to prevent collaboration by groups of students. It no doubt also tends to weight the situation in favor of the expression of public attitudes. Vetter has found definite evidence of a difference in the manner of checking attitude scales in the classroom as compared with solitary checking (72,p. 172) . Another variable factor is the personality of the instructor in charge and his relations with his students. Conklin and other investigators have tried to get a truer measure of private attitudes by "springing" the questionnaire upon the students. (18) Symonds also has emphasized the value of a quick, impressionistic, first response. (59)
When attempts are made to discover the overt behavior or the motives of individuals, it is often desirable to state behavior and motives as viewed by the subject himself. The customary rationalization employed by the student in defending his action may be included in the question in order to induce the student to recognize and acknowledge his own behavior. This device was employed in the Syracuse Reaction Study, for instance, in obtaining admissions of cheating in examinations. Meier has also used it in ascertaining motives in voting in the presidential election in 1924, a study in which all statements of motive were phrased in the same terms as those employed by the particular candidate. (42) Coolidge economy, for example, as presented by Republican publicity was another matter as presented by LaFollette.
The principle of indirection has been used by a number of investigators in attitude research. The very nature of some attitudes does not permit of direct questioning. The attitude (or perhaps it would be better to call it "trait") of fair-mindedness, studied by G. B. Watson and Symonds, must be ascertained indirectly, since one can scarcely be expected to have good insight into one's own prejudices. Form D of Watson's test illustrates this procedure. (74) Fifteen situations were listed, and the subjects were asked to approve, disapprove, or take a
(373) neutral position concerning the behavior described. Each in-stance in
the test was paralleled by one or two other instances presenting the same type
of situation. For example, in one item the account of an unwarranted search of a
suspected radical headquarters is given ; while later in the test there is
described a similar raid upon the offices of a large business corporation
suspected of dishonesty. Individuals who approve of this procedure in the one
situation and condemn it in another are considered to have shown prejudice.
Several examples from the Syracuse Study have previously been cited showing how
information may be elicited in the study of attitudes with, out throwing the
subjects too much upon their guard.