Experimental Social Psychology
Chapter 11: Social
Attitudes and Their Measurement
Gardner Murphy and Lois Barclay Murphy
ALL the difficulties which beset us in the exact study of personality inevitably trouble us in the study of attitudes, for attitudes, of course, are a part of personality. Yet the difficulties are even greater than they seem. An "attitude" is today generally defined as a "tendency to" or "verbal substitute for" overt behavior; but if the overt behavior does not happen, how can we be really sure that either the tendency or the thing verbally symbolized has any existence ? The final answer is not easy, nor, we believe, can it as yet be given. We shall regard attitudes here as verbalized or verbalizable tendencies, dispositions, adjustments toward certain acts. They relate not to the past, nor even primarily to the present, but, as a rule, to the future. Sometimes, of course, it is a hypothetical future. "Do you believe in . . . ?" is apt to mean, "How would you like to have so-and-so happen?" The "attitude" is primarily a way of being "set" toward or against things. Moreover, if attitudes are regarded as verbal substitutes for acts, the term must be used broadly. If an act, for example, is irrational, its corresponding attitude may be irrational. And since opinions are certainly, at least in relation to social behavior, verbal symbols for incipient actions, they will be included.
A few words, first, about sociological methods of approaching attitudes. Sociological variables, such as prosperity and urbanization, are much easier to bring into relation to attitudes than are most of the distinctly psychological variables. The statistical interrelations between crop failure, visits to towns, habits of newspaper reading, and so on, have already been hinted at. The
( 616) statistical pitfalls here are so numerous, however, that the derivation of sociological laws themselves is most difficult, quite aside from any questions of psychological interpretation. Willoughby, for example, correlates the subscription rate to the Nation per 100,000 native whites, with such variables as taxable incomes per thousand native whites and per cent urban population per state, and gets correlations of + .51 +.07 and .48 +.08, respectively. Whatever goes into the attitudes expressed in subscription to the Nation is related, then, more with prosperity than with poverty. Similarly, the correlation is striking between prosperity and suicide. Just how much information such data can give as to the laws governing attitudes can be safely set in the neighborhood of zero, except in special cases (such as those of Rice, described above, page 28 ff.) in which relatively homogeneous material is analyzed in relation to one factor at a time, logical analyses both preceding and following the interpretation of the correlations.
Another definitely sociological method is the large-scale comparison of case histories and of diaries, letters and records of interviews. Yoder's conclusion  that Iowa laborers are in general extremely hostile to "radical" leaders is, for example, sup-ported by such an abundance of materials of the type described that it is doubtful whether the giving of an attitudes test would add anything to general knowledge. To be sure, it is true that individual differences are more accurately revealed through the psychologist's methods than through sociological methods; but the question is, of course, whether the precise extent of individual differences is what we want to know. If we are sociologists, probably no; if psychologists, probably yes.
Another sociological study is Lundberg's comparison of the five most radical and the five most conservative counties of North Dakota, radicalism being calculated according to degree of support of the Non-Partisan League. The conclusions were that such radicalism was definitely related to soil and rainfall, that it was strongest in the newer and less settled communities, and in those
( 617) in which there was a larger percentage of immigrants and new population suffering under economic disadvantage. Similar data were obtained for Minnesota. Here, then, is a quantitative analysis of social causes of radicalism. Yet we do not see the individual farmer, nor hear what he has to say. For this kind of information we must turn to psychological methods.
One of the first things to do in constructing a real attitudes scale is, of course, to make sure of an order in which attitudes can be arranged from one point to another. No scale can really be called a scale unless one can tell from a given attitude that an individual will maintain every attitude falling to the right or to the left of that point (depending on how the scale is constructed). No one could use a ruler on which one could not tell whether the point marked 7 would fall between 6 and 8 or whether it would have a capricious preference for some other point on the instrument. If we have a genuine scale of attitudes toward compulsory military service, it is necessary to assume that every person favoring universal military service should also favor military service for persons conscientiously opposed to such service; the former proposition includes the latter. In the same way, a person who favors three years or more of military service ought certainly to favor two years of service if the choice is between that and no service at all. A person who believes that any insolent Negro ought to be lynched should believe that a Negro murderer should be lynched; a person who believes that half the population of the United States is feeble-minded ought also to believe that many of them are stupid. The question really is: What groups of attitudes are there which actually fall into a scale arrangement? Unless we arbitrarily define attitudes in such a way that they simply must, like the victims of Procrustes, fit themselves into a linear scale, our task is to find what attitudes there are in the measurement of which really valid scales can be constructed. This limitation of many alleged scales will be apparent at many points.
It has further been assumed, in relation to most tests of atti-
( 618) -tudes, that the final validation must come from rating scales—at least, that no test scores are worth much which are not correlated as high as + .60 or + .70 with the average ratings of close associates in respect to the attitudes measured. There are, however, ways out of this predicament. G. W. Allport, among others, has emphasized the desirability of validating all tests by applying them directly to persons who, in their actual conduct and social commitments, stand at opposite poles in respect to given issues (and, when practicable, to persons taking overt positions intermediate between these). Thus a scale of attitudes on economic radicalism and conservativism would make use of a group of Left-Wing Union leaders at one end, and of representatives of the National Association of Manufacturers at the other. A "radical" attitude could be defined simply as the average attitude of these labor leaders on a given issue; the employers' attitudes would be the kind that one could call "conservative." Intermediate steps could be found by taking the attitudes of leaders in the American Federation of Labor. Attitudes toward the authority of the Bible might similarly be validated by reference to the expressed opinions of evangelists and leaders of revivals, of Northern Methodists, of Unitarians, of "Humanists," and, perhaps, of the American Association for the Advancement of Atheism. It would be absurd to call these five steps of attitude toward the Bible a "scale"; not only would the truly representative expression of opinion be hard to find for each group, but if it were found the step-intervals would remain unequal. Nevertheless, with all its difficulties, the plan of validating in terms of an actual stand taken, or commitment made, is worth very serious trial.
A study in which this method of validation is combined with the usual rating-scale validation is Porter's Student Opinion on War. A thousand students indicated on a five-point scale the measure of their agreement or disagreement with each of about 150 brief statements about war, such as, "War is always wrong"; "In case of war, those who continue to oppose it should be im-
( 619) -prisoned"; "War is justified when its purpose is defensive"; "Armaments are the only sure guarantee of peace"; "Modern wars have been caused more by moral principles than by commercial questions." The questionnaire covers Alleged Causes of War; Alleged Results of War; Eliminating War; The Right or Wrong of War; Patriotism and Conscience; and American Action in the Immediate Future.
"For the purpose of standardization, 100 validation tests were obtained from persons whose position on questions involved was overt, and on which, in each case, from three to thirteen ratings by acquaintances had been secured on a scale of o to ro (extreme pacifism to extreme militarism). Validation tests were used only if ratings agreed within a span of three steps on the scale of eleven." These 100 individuals now took the test, and (after throwing out items which did not appear to be of value in differentiating attitudes) the score to be assigned to any given attitude is the average rating received by all the individuals who held that attitude. A large number of college groups show "nor-mal distributions" of attitude toward war, the mode lying at the mid-point on the scale in the great majority of the groups. Three denominational colleges make a modal response slightly toward the pacifist end of the scale; a state university R.O.T.C. class made a modal response slightly inclined toward the militarist end, but none of the variations are at all striking. A group of adult reserve officers make a distinctively high "militarist" score, and a group of conscientious objectors and student pacifist leaders make a definitely "pacifist" score. All the lines of evidence appear to converge and to justify the belief that the scale does actually indicate the amount of reaction against war. The scale not only serves the purpose of differentiating individuals but yields several generalizations of interest. Religious influences appear to play a very considerable part in the total pacifist score. Religious conferences of students, such as the Indianapolis Student Volunteer Convention of 1923, seem to have had a marked effect. Average scores of Roman Catholics on the scale were 5.4;
( 620) of Episcopalians, 5.1; of Baptists, 4.4; of Congregationalists, 3.7. Children of engineers averaged 5.4; of ministers and missionaries, 3.7. Another factor of importance is "social and economic liberalism," whether inspired by the faculty or by student organizations. The R.O.T.C. does not appear to exert a very great influence.
The most elaborate and mathematically sophisticated efforts toward the measurement of attitudes are those of Thurstone and his collaborators. As Thurstone explains, the "provocation" for his steps in this direction was Allport and Hartman's construction of a scale for the measurement of attitudes on such public issues as prohibition and the League of Nations (see page 665ff.). They had had six judges arrange in rank order a large number of written opinions on prohibition, distributing these from the "driest" to the "wettest" opinions, and had then calculated the frequency with which each of thirteen opinions on the prohibition question was held among a large number of students, and had asserted that the distribution of opinions conformed to a normal frequency curve. It will be evident that such a statement presupposes the equality of the steps between the various opinions. Now statistically no such assertion is warranted unless a rather elaborate procedure is followed in determining just how much drier one opinion is than another, just how much more paternalistic one is than another, etc. There must be, not six, but a very large number of judges, and these judges must be made to determine the relative positions of a large number of opinions, varying from instances in which the decision is so clear that all judges agree, to instances in which the issue is so difficult that the judges divide evenly. On this basis, Thurstone pressed into service the methods of psychophysics, which for many years have been available to the statistician who is interested in the capacity of persons to discriminate large or small differences. A unit, such as on; gram, may in a given psychophysical experiment be sufficient to permit discrimination to a point represented by the fact that 75 per cent of the judges make the distinction in the right
( 621) direction (as compared with objective measurements). A difference of two grams might serve for the basis of a correct judgment in nearly all judges, the frequency of error decreasing, of course, as the objective difference in weights increases. Conversely, Thurstone contends that two stimuli may be considered to be actually different by a quantity determined by the percentage of judges who can perceive the difference. If, for example, 75 per cent of the judges agree that a given opinion about the liquor problem is drier than another given opinion, the two opinions actually differ by one unit. If, now, Allport's thirteen opinions are compared, not by six, but by two hundred judges, it is found that in some instances the relative dryness of two opinions is agreed upon by less than 75 per cent of the judges, so that the two opinions ought to be rather close together on the scale, whereas the relative position of other opinions which have been regarded by Allport as adjacent are agreed upon by almost all the judges. On the basis of the findings of two hundred student judges, Thurstone reconstructs Allport's original frequency curve and shows that its shape is profoundly altered by the more careful scaling method. Proceeding to other problems than those with which Allport was concerned, Thurstone has drawn up a scale for the measurement of opinions regarding the seriousness of various crimes, each crime from murder to vagrancy being assigned a given scale value, based upon the opinions of two hundred students when asked to compare each of twenty offenses with each of the others. The difference, for example, between burglary and arson may be stated to be just so many times as great as the difference between arson and vagrancy, in terms of the degree of definiteness with which a large group of persons render their judgments.
Following Thurstone's plan, scales are now available for measuring various other attitudes. It must, of course, be remembered that such scales do not presuppose that all judges would agree as to the position assigned to any given attitude. The whole merit of the scale is that it presents in quantitative form a consensus
( 622) of opinion in which arguable points automatically separate them-selves from non-arguable points, and the sole authority is the degree of evidentness of a distinction to a large number of people. In his study of nationality preferences, for example, Thurstone presented to nearly 250 subjects a series of paired comparisons in which the individuals were told to choose by underlining "the one nationality or race of each pair that you would rather associate with." The order for these American students was almost identical with that found by other methods (see pages 630 and 638), namely: American, Englishman, Scotchman, Irishman, Frenchman, German . . . Italian . . . Greek . . . Japanese .. . Hindu, Turk, Negro. But Thurstone is concerned, of course, not merely with ranks, but with distances.
Defining the method in somewhat greater detail, Thurstone and Chave  present data from 300 subjects who classified 130 opinions about the church into eleven piles on the basis of equal appearing intervals. The median position of each opinion was taken as "its scale value," and the interquartile range (showing how widely the judgments differed) was taken as a measure of its "ambiguity." The opinions were then submitted to another group, and those statements which were endorsed about as many times as those having widely different scale values were thrown out. The 130 opinions were thus reduced to 45, the reliability of which, by the split-half method, was .92. The scale was then tried (as in Porter's study, see page 619) on groups who had overtly committed themselves to a position, and validation was secured.
Another method of validation of Thurstone's is best exemplified by Stouffer's study  of attitudes toward prohibition, which merits detailed description.
The purpose of the experiment was to compare a statistical and a case history method of attitude research. The subjects were 238 students who took the H. N. Smith test of attitudes toward prohibition (constructed according to Thurstone's method). The students also wrote anonymous accounts of their experiences,
( 623) feelings, and opinions with respect to prohibition laws and to drinking liquor, from childhood to the time of writing. Each account was about a thousand words long. The case histories were read by four judges working independently, and were rated on a five-inch graphic rating scale to show the extent to which the attitudes were favorable or unfavorable toward prohibition laws. The judges were selected by a committee of the faculty of the Department of Sociology on the bases "of knowledge of the theoretical literature on attitudes, of experience in interpreting case materials, and of insight into human experience." They were told to use "whatever concepts of attitudes on prohibition they chose, to set their own standards of favorable or unfavorable, and to try to judge all papers by the same standards." The average of the six intercorrelations of ratings of each judge with ratings of each other judge was +.87, the range being from +.83 to +.89. The ratings by each judge were expressed in standard scores (deviations from the mean of his rating, divided by the standard deviation).
The validity coefficient, measuring the extent to which scores on the Smith test were correlated with composite standard scores on the case history ratings by four judges was +.81, which be-came +.86 when corrected for attenuation. The validity coefficient of +.86 was checked in several ways, all of which tended to confirm the apparent fact that attitudes as measured by the test and attitudes as measured by the case history ratings are quantitatively very much the same.
First, the students rated their own attitudes toward prohibition laws on a graphic rating scale similar to that employed by the judges. They rated themselves twice, with two days intervening, and the two ratings were combined, giving equal weight to each. The correlation of the self-ratings with the Smith test scores was +.80. The correlation of the self-ratings with the composite ratings of the four judges on the case histories was the same, +.80. "Unless spurious factors entered to account for the agreement between the findings of the statistical and case-history
( 624) method, one would seem justified in concluding that, whatever it may be that the two methods were measuring, each measured substantially the same thing."
The judges asserted that they used the same general definition of attitude that is used by the layman when he discusses "so and so's attitude about prohibition." The likelihood that the concepts of attitude of the judges corresponded closely to a common-sense usage tends to be confirmed by the fact that the ratings on a sample of 99 papers by two laymen, the superintendent of the Illinois Anti-Saloon League and the secretary and director of the Illinois Association Opposed to Prohibition, yielded about as high correlations with the ratings of the four judges and with each other as did the correlations of the four judges among themselves.
Stouffer urges caution in interpreting the validity coefficient, stating, "It means simply that under a given set of experimental conditions, a correlation of +.86 was found between attitude scores on a test constructed by statistical methods and attitude ratings by judges, selected for competence, who used case history methods of interpretation." If more data could have been obtained by interviewing the students or their acquaintances, the coefficient might have been lowered. If attitudes should be de-fined in some logical way deviating materially from the modal group of meanings, the validity coefficient might be different. The coefficient might even vary using the same experimental set-up, depending on the experimental situation itself. But, in sum: "(1) The validity coefficient of +.86, in conjunction with the test reliability of +.94, lends confidence in the Thurstone method of measuring attitudes. (2) The validity coefficient of +.86, in conjunction with the case-history ratings' reliability of +.96, lends confidence in the objectivity of the case history method."
To the objection that such methods are artificial, it must be replied that any quantification of an attitude involves schematization, the ultimate value of which for scientific purposes is not to
( 625) be prejudged, but only to be estimated after much more research (especially on the genesis and the interrelation of attitudes). To the further objection that such scales are only valid for the kind of people who themselves acted as judges, it must, of course, be admitted that gross differences in education, social status, etc., would make what is a small difference, a large difference, and vice versa; but that a statement as to the kind of judges used in standardization is an integral part of any such research instrument, and that the decision as to the applicability of any such instrument must always rest with the person using it. A third objection is perhaps more difficult to meet. After all, attitudes in Thurstone's sense are verbal behavior, or, to be exact, are paper-and-pencil behavior with respect to verbal statement stimuli. Even the case histories are but verbal behavior. The relation between such paper-and-pencil behavior and actual social conduct can only be defined by very laborious research. As we shall see, how-ever, the problem is not hopeless, and the clear mathematical study even of mere verbalizations and of the interrelationship between such verbalizations is of the utmost importance in our present chaotic state of inquiry regarding attitudes and opinions.
Though full of limitations from the standpoint of technique, Sorokin's attempt  to check up on the relation between verbal and overt behavior is worth mention. He asked six sections of a sociology class to contribute to three causes: (1) to buy diagrams and calculating machines for their own departmental use; (2) to help three brilliant students whose families were impoverished by floods so that they were forced to leave college; (3) to help Chinese and Russian students suffering from famine. The money vote which was immediately taken showed that both the amount to be given and the number voting to contribute to each cause were in direct relation to the nearness of the appeal, i.e., to their own classes first, to other students second, and to Russian and Chinese students third. Later, when a true-false test was given on a phrase declaring that we should be equally ready to help all, no matter where they may be, most of the students marked
( 626) the statement true. It will be noted that discrepancies of this sort bear a rather meager relation to the kind of discrepancies between verbal and overt behavior with which economic and political life chiefly deals; but even this kind of check-up on the verbal response cannot be overlooked.
But since the point about the "merely verbal" character of much attitude research has been rather heavily emphasized—and, in fact, quickly granted by most investigators—it is worth while to emphasize the relations already shown to exist between con-duct, on the one hand, and verbal report of attitude, on the other. In Porter's study of students' attitudes toward war, student pacifist leaders—that is, those who had overtly identified themselves with pacifist activities—scored near one end of the scale, while officers in charge of R.O.T.C. scored near the other end. Neumann  applied his test of international attitudes to forty persons definitely identified with movements or activities of genuinely international character, such as business men doing a large volume of business overseas, members of the Communist party, and leaders in international religious movements. Here it was found not only that the paper-and-pencil behavior agreed with overt conduct, but that there was not a single exception in test behavior to the predicted answers. After all, a man's categorical agreement or disagreement with a rather strongly stated opinion about Chinese, or Jews, or Communists, or Rotarians, is in everyday life regarded (if the man be sincere) as a significant part of his behavior. There seems to be no reason why this behavior should suddenly become non-significant when it is made the subject of careful inquiry.
We pass now to some practical uses of attitude tests.
Among the many recent studies of interracial attitudes, the most comprehensive is G. B. Watson's Orient and Occident,
( 627) prepared for the Research Committee of the American Group, Institute of Pacific Relations, before the 1927 Conference of this Institute, in Honolulu. An elaborate questionnaire was prepared, with the aid of both oriental and occidental students, on the interracial attitudes prevailing between Americans and Pacific peoples. Some of the questions have to do with "How you feel," the subject checking, among a number of phrases, that one which most adequately expressed his first response to a stimulus phrase. The phrase, Japanese Exclusion Act, for instance, is followed by the choice of responses: (I) Insult; (2) Unfortunate; (3) Necessary; (4) Desirable; (5) Doesn't go far enough; (6) Must be undone. Another section of the inquiry begins with the phrase, "What you think," the subject responding to each of a series of sentences by checking symbols standing for "true"; "prob-ably or partly true"; "in doubt" ("divided," "open question"); "probably or partly false"; "false." The subject must characterize by one of these symbols the sentences, Japan is preparing to fight the United States; or an American doing business in China ought to be willing to abide by Chinese law, even though he considers it inferior to that of his own country, etc. Some of the issues deal with missions, disarmament, and other matters not exclusively associated with oriental-white relations. The questionnaire also gathers information as to age, sex, race, occupation, schooling, church membership, reading habits, and other important personal data applying to those who answer the questions.
Ten thousand copies were sent out, and about one-third of them were returned. In some groups, such as high school classes, 100 per cent of those given the questionnaire filled it out, whereas in some of the others, such as farmer groups, the number responding was very small, so that the factor of selection seriously jeopardizes the results. The amazing range of social groups studied is indicated by a few samples: University of Chicago class in European history; students of naval science, University of Washington; Methodist Young People, Oak Park, Illinois; business girls, Houston, Texas; Palo Alto (California) Kiwanis Club; .Ohio clothing workers; Richmond (Virginia) business men;
( 628) New York Urban League (for Negroes) ; Farmington (Missouri) housewives; prisoners, Bridewell Jail, Chicago; Methodist Episcopal Church, Waterloo, Iowa.
Group profiles are presented in which the following appear: attitude toward Chinese; attitude toward Chinese nationalism; attitude toward Japanese problems; attitude toward Philippine independence; attitude on United States policies, etc. In each of these major divisions of the data, attitudes are reckoned in both directions from a zero point, deviation to the right indicating favorable responses (e.g., to the Chinese), and those toward the left indicating unfavorable responses. The profiles of the entire group of over 3,000 people are separately compared with each of the following: some American members of the 1927 Conference of the Institute of Pacific Relations (which had a very small "unfavorable" score toward Chinese, very low "unfavorable" and high "favorable" score toward Japanese); Protestant missionaries (very low "unfavorable" score toward Japanese, high "favorable" score toward Japanese) ; Kiwanis Club in small California city (high "favorable" score toward Philippine independence); employees of a New York bank (high "unfavorable" score toward Chinese, low "favorable" score toward propositions of an anti-imperialist nature). It will be noted that a high favorable score does not invariably mean a low unfavorable score, although in most cases the two do go together. In order to get larger groups for comparison, many small groups scattered through many states were lumped, and these permit the following generalizations: unfavorable attitudes toward the Chinese are especially prevalent among the labor group as a whole, and least prevalent among the housewives and women's clubs; the composite group of business men shows a very small amount of agreement with anti-imperialist propositions; the high school students are very much below the score of the entire group in responses favorable to Philippine independence. All these data are suggestive, and those obtained from groups of students are probably fair samples of the larger groups from which they are taken. But the ordinary statistical precautions for the study of the reliability of differences
( 629) hardly need to be labored here, on account of the difficulty of estimating sampling errors.
No matter how serious such errors may be, however, a great deal of interest nevertheless attaches to the fact that the geographical differences which are often so much emphasized fail to appear in the results. College students of the eastern, southern, mid-western and Pacific states differ but little; "California business men are not clearly different from those of the East"; and some differences which do appear (e.g., the interest of the southern group in foreign missions) do not appear to be the result of interracial contact. Economic differences seem to be significant. "California Labor is more like Ohio Labor than it is like the business men of California."
Some of the items in the questionnaire had to do with matters of information. In general, information scores on relevant issues were associated with favorable attitudes toward the Japanese (r = +.82 +.04) and toward Chinese nationalism (r = +.70 +.06). The influence of information—or perhaps we should say familiarity—is shown rather dramatically in the fact that a favorable response toward chow-mein is correlated +.67 with the total favorable reaction to all questions relating to China (the correlation is based upon lumping together ten groups of very diverse questions so that the figure probably ought to be reduced slightly). As Watson points out, this "sounds too journalistic to be a matter of sober fact." Here, as elsewhere in the problem as to the relation between information and attitude, we are dealing not with a simple and obvious causal relation but with the dependence of both of our correlated variables upon a variable (or a group of variables) which is not directly measured.
International and interracial antagonisms have, then, lent them-selves easily to measurement—even, indeed, to experimental study. One of the most striking attempts to state results in quantitative form is the "social distance" method devised by Bogardus, and modified and applied by many others. The test itself will show most clearly what social distance means. The subjects are asked to respond to the following sort of questionnaire:
DIRECTIONS: According to my first feeling reactions I would willingly admit members of each race (as a class, and not the best I have known, nor the worst members) to one or more of the classifications which I have circled.
Bogardus's social-distance test has now been put to so many uses and with such interesting results that it requires rather extensive discussion. In one compilation, 1725 native-born Americans from all over the United States are studied. Most of them had college or high school education. The great majority are of north European ancestry. The definiteness with which the test acts, for most subjects, as a "scale" is shown by the following figures for this group (stated in percentages) :
It is interesting to note that a Negro group of over 200 subjects shows the following rank order for nearness: Negroes, mulattoes, French, Spaniards, English, Canadians, Mexicans, Americans, Hindus, Japanese . . . Chinese . . . Turks. Again, 178 native-born Jews showed the following ranking: German Jews, Russian Jews, English, French, Germans . . . Spaniards . . . Japanese ... Chinese . . . Hindus . . . Negroes.
One of the most interesting parts of Bogardus's work is the study of changes in attitudes during recent years. These have been tabulated as changes to a less favorable or a more favorable attitude. Increasing antipathy toward the Japanese in recent years has occurred in nearly 20 per cent of a group of 508 Americans. Increasing antipathies toward Negroes appear in 13 per cent, toward Russians, 7 per cent, and so on. Favorable and unfavorable changes are "balanced" in one of Bogardus's tables, and an algebraic "adverse balance" puts the Turk at the greatest disadvantage of all. Increases in friendliness are nevertheless very striking in many cases; there are about as many favorable as unfavorable changes, for example, in relation to both Japanese and Negroes.
Bogardus has further proceeded to analyze his data according to sectional differences, showing how attitudes toward Negroes on the part of southern whites compare, for example, with attitudes of whites in the southwest toward Mexicans. A considerable quantity of case material is offered to make these attitudes and changes of attitude intelligible. The great importance of individual conditioning incidents is stressed, although it is recognized that the causes for attitudes are usually "indirect, subtle and many-sided." Variations with age and sex are also reported. The comparison of adults with adolescents presents many curious contradictions, adolescents being more "friendly" than adults toward a given group at one point on the scale, and then reversing them-selves at the next point (for example, comparing the "neighbor" column with the "occupation" column). Sex differences are remarkably slight, and this holds not only for attitudes taken at
( 632) the moment but for changes in attitude. Religious differences appear in general to be of but little consequence.
There is, of course, no effort here to make the step-intervals equal, and in our own experience we have been impressed by the frequency with which steps on the scale are skipped—cases, for example, where the individual may be willing to admit members of another group to his street as neighbors but not to employment in his occupation. Even the skipping of two or three steps is to be expected occasionally. Nevertheless, Bogardus's own data indicate that there is, in general, a progression in social distance (page 63o) sufficiently definite to permit our saying that if an individual is refused from a given class, he will usually be refused from all "nearer" classes. The method, when used with over 1700 American students, showed in fact that the social distance was exceedingly small where the groups concerned were closely related or believed themselves to be closely related. The social distance of Canadians from Americans is practically nil; that of Germans is larger; that of Chinese is very great. It is remarkable to note that in some of the work done with this scale, the social distance of the Turk is so great. It is unnecessary to go farther than history texts and war-time atrocity stories to explain this.
As a supplementary method, Bogardus in another investigation has asked groups of students to rank a large number of national or racial groups in order of personal preference. The results appear consistent: In one experiment with 248 students, the races against which the greatest antipathy was expressed were as follows: (1) Turk; (2) Negro; (3) mulatto; (4) Japanese; (5) Hindu, etc. Against the following groups no antipathy scores were recorded at all: Canadians; Danes; Dutch; French Canadians; Norwegians; Scotch.
It was noted above (page 616) that sociological methods usually pay less attention to the individual than do psychological methods. Yet the absence of a sharp distinction is apparent in the following. In many ways the most satisfactory study of interracial attitudes yet published is that of Lapiere. Relying upon
( 633) interview rather than printed questionnaire, and concealing the purpose of his inquiry, he gathered data from hundreds of per-sons in France and England as to their attitudes toward colored people. In France he took occasion wherever he happened to be on his travels to put to 428 people some form of the question, "Would you let a good Negro live at your home ?" The question was always introduced as part of a conversation which had been shaped in such a way as to make the question natural and inoffensive. A stereotyping of replies could, of course, not be expected; these were classified as "without prejudice," as "doubtful," or as "with prejudice." The results are:
|No. of Cases||With Prejudice||Doubtful||Without Prejudice|
The striking increase in anti-Negro feeling in cities as against country districts is in part explained, Lapiere points out, by unpleasant memories dating back to the conditions of dock labor at seaboard cities. Lapiere proceeds from these gross totals to an analytic study of the relation of racial antipathy to economic status. Since no more accurate basis of classification was possible, he classified his subjects in terms of the places where he met them (for example, in a first-class railway carriage, in a cheap café, etc.). The detailed figures are striking:
|Classes||Without Prejudice||Doubtful||With Prejudice|
Anti-Negro feeling, in other words, is exceedingly rare in the middle and lower classes, whereas it predominates among the upper-class subjects.
As a check upon these personal interviews, Lapiere hit upon the idea of inquiring as to the policy of hotel managers and inn-keepers. As he points out, the hotel man cannot consider himself an entirely free agent; he can scarcely refuse guests whom the community would see no reason for excluding, nor can he, on
( 634) the other hand, admit guests whose admission would offend most other people. Of thirty-one hotel proprietors who were questioned as to their willingness to admit Negroes, twenty-four replied affirmatively (and Lapiere saw, as any one can, Negroes being courteously entertained in French cafés and hotels). Of the seven managers who said that it was contrary to their policy to admit Negroes, the majority were among those catering to foreign visitors, especially Americans. There is, then, very considerable evidence here that the great majority of Frenchmen feel no resistance either to the entertainment of Negroes at hotels, or even to the admission of "good" Negroes to their homes.
Striking indeed is the contrast across the English Channel. Here Lapiere continued his inquiry, conversing on race relations with several hundred English men and women. For reasons which he does not fully explain, he found it necessary to alter the form of his question, so that the resulting data must not be regarded as in all points comparable. In various parts of England he asked people of various social classes some form of the question, "Would you let children associate with those of good colored people?" If his subjects had children of their own, the question was, of course, in the form, "Would you let your children," etc. The results, according to social classes, are as follows:
|Classes||Without Prejudice||Doubtful||With Prejudice|
Here there appears not only a very striking contrast with the French results, but a complete absence of that socio-economic differentiation which was found among the French. The racial antipathy recorded here seems to be as great for the lower as for the upper classes; in fact, it seems to be strongly marked among the lower-class group. Among twenty hotel managers questioned regarding their policy toward colored people, sixteen
( 635) replied that they would exclude them. In general, Lapiere found but little difference between attitudes toward Negroes and those toward natives of India; antipathy seemed to be shared toward all dark-skinned people.
The obscurity of the origins of these striking national differences is recognized. Lapiere contents himself with the very mod-est conclusion that racial antagonism is not innate. Actually, little has been done (compare, for example, Green's data, page 646) to study the origin of racial prejudices in Great Britain, though a fairly satisfactory picture can perhaps be drawn in terms of the economic and military history of the British Empire. It is very much to be doubted, however, whether this extreme and dramatic difference between France and England could have been wholly foreseen on historical or sociological grounds. Whether any of the existing contrast is due to the different form of the question is uncertain. It seems possible that Lapiere, anticipating greater prejudice in England, intended to ask a less provocative question. If this was his purpose, he probably did not succeed.
The question now arises as to the practicability of getting more quantitative data as to the basis for such antagonisms in the life histories of the individual subjects. The most systematic study here appears to be that of Diggins  His problem is in the form, "To what extent does greater familiarity with individuals from various national groups affect opinions regarding these groups ?" His subjects were 87 foreign students and 24 American students in International House in New York City, 25 students at Columbia University, 73 students at the University of Indiana, and 78 students at Occidental College, Los Angeles. The students at International House were reached in person. Those at Columbia University received a circular letter; those at Indiana and Occidental College were reached through a classroom experiment. The task assigned to the subjects was to rate each of ten national groups—Chinese, French, Greek, Japanese, Russian, English, German, Hindu, Norwegian, American—in respect to the follow-
( 636) -ing attributes: (I) Art (including painting, music and sculpture); (2) Industrial progressiveness; (3) Personal preference of the subject, considering such factors as meeting members of these groups socially, having them as friends and companions; (4) How well the subject would like to live in a neighborhood in which all families were of this nationality. After ranking the ten groups in respect to these four characteristics, the subjects recorded their own sex, nationality, and the countries which they had visited or lived in (stating the length of such visit or residence) ; the groups of which they knew at least three native-born members ("in the sense of having seen them on the street ... or knowing them well enough to speak to ..."); the national groups of whom three or more individuals were intimately known to the subject (well enough to "talk to them often and know their personal characteristics").
The merit of the method lies in the fact that every national group must be assigned a definite rank. It is impossible, on grounds of general "humanitarianism" or general "nationalisms" to rate all groups high or all groups low. It is now possible to compare the above data with regard to personal experience of foreign peoples with the ranking in respect to the various questions. Counting 1 as the most preferred and 10 as the least preferred group, the following figures appear with reference to the influence of travel in foreign countries. Comparison is made with the answers to question (3), which is the one about personal preference. Each group had to rate all ten groups.
|Travel in Foreign Countries||No Travel||Less than One Week||One Week to Six Months||Six Months or More|
|International House Americans||6.44||4.75||4.87||4.21|
Whatever the causal relations may be, there is a fairly clear tendency to rate a group higher, the longer one has been in contact
( 637) with the group on its own soil. The result is even more striking when we compare the number of intimate acquaintances with the relative standing of the groups. The average position of all those groups among whom the given subject has no intimate acquaintance is 6.14; for those among whom there is one acquaintance, 5.07; two acquaintances, 4.50; three acquaintances, 3.19; four acquaintances, 2.65 (figures relate to entire group of subjects). This progression appears equally clearly among the foreign students, among the International House Americans, and among the Indiana and Occidental College students. These data may also be regarded as a measure of neighborliness, since the correlation between the group standings in reference to questions (3) and (4) is +.95 +.02.
Diggins proceeds from this to a more elaborate attempt to measure prejudice in terms of a spurious consistency in the standing of the groups in reference to the four very different questions which were asked. There is no special reason why a nation whose art is superlatively fine should be one which is outstandingly progressive industrially. Any excessively high correlation in ratings would be reminiscent of the "halo" error so well known in ratings of individuals. The consistency index is therefore computed, from which we can learn to what extent groups are judged in the "halo" fashion. "Our theory is that prejudice lurks behind a consistent rating of varied factors"; a high consistency index implies prejudice. Consistency indices are worked out for the various groups of subjects who act as judges. The consistency indices are in the following rank (from low to high) : foreign students, International House Americans, Columbia, Indiana, Occidental College. The foreign students, as judged by this method, are the best judges—that is, they make the best discrimination between the various national groups, being willing to recognize one "good" trait in a group simultaneously with the presence of some supposedly less desirable trait. The Indiana group, which is probably relatively little informed about the national groups in question and has had but few contacts, makes
( 638) the poorest discrimination; the students at Occidental College, however, stand only a short distance away.
Diggins's work is unfortunately impaired, as he himself points out, by the inclusion of self-ratings, or, rather, the ratings of one's own group. Fortunately the outlines of his results are not disturbed by this consideration; a case might have been made, perhaps, for the importance of using ratings of one's own group in comparison with others, but the data should, of course, be disentangled. That a striking result remains even after the removal of this possible source of error is shown by the fact that the rank difference correlation between preference and familiarity (familiarity measured by assigning a certain weight to intimate acquaintances, slight acquaintances, travel, and familiarity with foreign languages) is +.82 +.07. It may be of incidental interest to know how the ten groups stand in the eyes of two groups of American students. For the Columbia University group, the preference ranking is (1) American, (2) English, (3) French, (4) German, (5) Norwegian, (6) Russian, (7) Greek, (8) and (9),Japanese and Chinese tied, (Io) Hindu. For the Indiana group, the first seven are identical with those given by the Columbia group, followed by (8) Japanese, (9) Hindu, (10) Chinese.
Those interested in the abolition of racial and national antagonisms are, reasonably enough, gratified at this evidence that antagonism is largely a function of lack of familiarity. The presence of students of various nationalities and races at the same institution has, of course, frequently been found a most effective method of wiping away even some of the most aggravated grievances. We must, however, turn to the other side of the story. A number of considerations clearly indicate that "familiarity," at least in the ordinary use of the term, has, as such, no clear implications for racial understanding or liking. The most direct approach to this is the study of the relations between whites and Negroes in the United States.
Economic and other sources of friction have of course always existed, and it has been a commonplace to point out within recent
( 639) years that as the Negro has moved north, especially to the Middle West, the relative friendliness which he found awaiting him some years ago has tended in many regions to give place to the most violent antagonism and discrimination—in fact, to a number of race riots involving wholesale misery. It may, indeed, be that greater genuine familiarity with the Negro is not involved here, for whites became familiar merely with certain aspects of Negro life. But usually knowing about people is one thing under conditions of regulated social life in which opportunities for conflict are reduced to a minimum, and quite another thing when, during periods of industrial chaos, newcomers are brought into active competition for jobs among another group. The latter can scarcely be expected to welcome a new competitor, especially one at the bottom of the economic ladder. The crying need for research on Negro-white antagonisms has, as far as we have discovered, resulted, as yet, in very little exact research.
From the point of view of method, however, Hunter's work  merits a rather detailed description. Herself from North Carolina, long interested in interracial attitudes, she arranged for the study of racial attitudes as they exist in New York City and in North Carolina. Eight groups were studied. In North Carolina: (i) 40 students at the University of North Carolina filled out a questionnaire in a psychology class; all were born south of Maryland and had spent all or most of their lives in the South. (2) A group of 40 students at Shaw University (a Negro institution) filled out a questionnaire (differing necessarily in some details from that given to whites, since the questions upon which an expression of antagonisms might be expected would, of course, not be the same for the two groups). In the case of these two student groups, all questionnaires were returned. (3) A white Protestant church group in a mill town. The 40 questionnaires were given out at a church meeting. Of the 31 which were returned, two were blank, the subjects saying that they were too opposed to the whole thing to fill out a questionnaire. The subjects included mill owners, operatives, and other residents of the town. This group is
( 640) the least homogeneous of all (a more intense antipathy to the Negro appearing among some of the mill operatives than elsewhere in the group). (4) A church group of Negroes. Forty questionnaires were sent to the minister; 24 of these were re-turned filled out. The four groups in New York City are as follows: (I) 62 undergraduates in a psychology class at Columbia College (three of the subjects were born south of the Mason and Dixon line). (2) The Intercollegiate Society, a group of Negro students and graduates; 29 questionnaires were given to members at a meeting and filled out. Three other questionnaires from Negro students in Columbia University were added. (3) A white Protestant church group. Of 40 questionnaires given out 39 were returned. There were some members of the "working" class in this group. (4) A Negro church group in Harlem. Of the 40 questionnaires, 31 were returned.
Since the questions dealt chiefly with points upon which a friendly or hostile attitude could be indicated—at least if the terms are used broadly—it is possible to compute a "good-will scale" for the comparison of the four white groups. This seems to us the least objective part of the work and may be summed up very briefly. The scores on the basis of a possible 100 are as follows:
|University of N.C.||49.0||13.0|
Turning now to the percentages giving various kinds of answers on specific questions, some striking sectional differences appear:
It will be noted that in some cases the difference between college students as a group and adults as a group is greater than between the northern and southern groups. In view of the small and selected nature of all of the groups, it is hardly worth while to compare these percentages for statistical reliability; it is only when the differences are exceedingly large that much importance can be attached to them. Considering this a preliminary study, the most important thing, perhaps, is to point out that there is, as far as we can tell from these data, no such thing as a gross "sectional attitude" as a whole. Differences between northern and southern white attitudes are highly specific and have to do with
( 642) issues upon which crystallized public opinion already exists. The question of eating with the Negro is one on which such a crystallized difference of opinion does exist. It will be noted, how-ever, that a majority of all four groups believe that some Negroes should receive a college education, and that there is no reason to believe that the group of southern white students is more hostile in this respect than is the corresponding northern group. Opposition to lynching under all circumstances is just as clear in the southern group as in the northern. The question arises as to the possibility that the differences between students and adults may lie in age rather than in education or in any other selective factor. Lumping together all the white adults, however, as a check on this point, it turns out that the 27 subjects who are 33 years of age or younger have a score of 43.1 on the "good-will scale" (S.D. 18.3), whereas the 43 subjects who are 34 years of age or older have a score of 49.0 (S.D. 19). As far as the data go, age probably makes no overwhelming difference.
Hunter then proceeds to analyze, as far as the subjects are capable of answering such questions, the causes of their antipathy toward Negroes: "Which of the following is the chief cause of any antipathy you may feel toward the Negro: (1) His lack of personal cleanliness; (2) His squalid home; (3) His economic competition with the white man; (4) His immorality and shiftlessness; (5) The color of his skin; (6) The preponderance in number?" The results (in number of cases, not per cent) are as follows:
|University of N.C.||38||19||6||7||22|
Seventy-two per cent of the total is made up by grouping together "personal uncleanliness," "squalid home," and "immorality and shiftlessness." It is scarcely maintained that these subjects, or any
( 643) others, know the real reasons for their antipathies. It is clearly important, however, as one step in understanding the matter, to obtain such self-observations as we may.
We can at leapt get beyond the cruder form of rationalization in the following question, "Is there a Negro whom you regard as a friend?" The replies here show a peculiar danger in assuming an undifferentiated hostility in one part of the country as opposed to liberality in another. The majorities of both southern groups answer "Yes."
|University of N.C.||39||28|
The comparison of the four Negro groups on the "good-will scale" shows but slight differences in the group averages and very large standard deviations, except that the college group in the North appears to be somewhat embittered as compared with the northern adults. The "good-will scale" (not comparable with the "good-will sc ale" for the whites) shows 33.7 for the Intercollegiate Society, S .D. 14.4; and 45.4 for the northern adults, S.D. 17.3. In reply to the question, "What incidents and situations make you most racially conscious?" segregation and the Jim Crow car appear as decidedly the most important among the specific cases given, whereas the somewhat more general terms, "discrimination," and "economic discrimination," appear likewise important. The southern university group is, naturally enough, the one which refers most frequently to segregation and the Jim Crow car. We doubt whether any figures could show more clearly than the following the crux of the problem of Negro-white relations:
|Intercollegiate Society||6||Columbia College||81|
|Shaw University||3||University of North Carolina||95|
|Northern Adults||0||Northern Adults||97|
|Southern Adults||23||Southern Adults||96|
The question is also asked of all the groups as to whether race antipathy is increasing, decreasing, or stationary. The answers here give no conclusive evidence, although among the northern groups (both races) there seems to be a tendency to emphasize decrease in antipathy more frequently than increase. A striking race difference comes out in reference to the question of the influence upon white people of the progress made by the Negro. All four of the white groups, by large majorities, believe that such progress has decreased antipathy, whereas the majorities of the Negroes disagree. At the close of the questionnaire all groups are asked what they believe to be the Negro's future, economically, politically, and socially. Very few in any of the groups, whether white or Negro, appear to believe that the Negro will attain equality with the white man in any one of these three fields. On the other hand, the majority of both races expect some Negro progress.
Some illumination with reference to the causes of racial antipathy appears in the following percentage table:
|University of North Carolina||30|
As we saw (page ''off.), no one yet knows what the Negro's capacity, as compared with the white man's, may be, but to rate him 50 per cent as intelligent as the white man is to assign him a lower place than any of the test evidence assigns him, and to count him only 25 per cent as intelligent is, of course, to express a conviction which even at its best could scarcely be expected to lead to racial understanding. The 25 per cent estimate reminds one of a remark occasionally heard, that the Negro must be treated like a child—an attitude which is related in no uncertain way to the bitterness of the northern Negro students as noted above. But it will be recalled that majorities in all the groups
( 645) studied are strongly favorable to Negro education—even to "college education for some Negroes."
Of course, the general question of interracial attitudes must yield sooner or later to the question of varying attitudes to different members of different races. Stereotyped responses about "niggers" or "gringos" play their important part in racial conflict. An important part, nevertheless, is played by personal relation-ships which cut through these stereotypes, resolving an otherwise unbearable situation into elements frequently human enough to make possible present-day coöperative scientific and artistic enterprises.
A study like Hunter's leaves one with the feeling that the concept, "contact means friendliness," is too general to help much in the present problem; whereas such a statement is valuable at a certain economic and cultural level—in the reduction of antagonisms, for example, between educated Chinese and educated Americans—it is no panacea. The causes of racial antipathy are legion.
How early do such antagonisms begin? The origins of these attitudes, as Lasker  has shown in such a detailed and exhaustive way, are usually in childhood experiences, among which a few dramatic incidents may stand out as significant. It must, however, be remembered that the kinds of incidents which will happen to the child and which will impress themselves upon him depend in large measure upon the attitudes of the parents, and that what would be speedily forgotten if the parents were in cordial relations with another social group may be reinforced and kept alive by parental pressure. To be sure, there are abundant evidences (compare Lapiere's study, page 632ff.) to show that such prejudices may sometimes arise in the adult period; but Lasker would doubtless repay quite rightly that if absolutely no hostile furrows had been formed in the child's mind, later experience would be powerless to produce such a strong effect.
Supplementary to Lasker's large array of case material on racial attitudes among American children is the recent ques-
( 646) -tionnaire study made by G. H. Green  among Welsh children, having as its purpose, first, to ascertain the presence or absence of prejudice against British, Americans, French, Germans, Chinese, Negroes, Russians, Spaniards and Italians; second, to find the origins of such prejudices, with emphasis upon school, newspapers, books, home, religious institutions, movies, and personal experience. It appears from this that the movies have had considerable influence on attitudes toward Chinese and Negroes, but that they have had practically no effect upon attitudes toward Americans (place of manufacture of the movies is not given, but there are surely a fair proportion of American origin). Green finds that books are in general more important than either school or religious institutions; obviously personal contacts played practically no part in the case of some of the races and nationalities named (cf. page 636).
It may be of interest to compare Keeny's study  of the origins assigned by nearly 400 students to their present attitudes on a variety of issues. No less than 717 specific experiences were reported which seem to have conditioned attitudes; they came from school (21 per cent), reading (21 per cent), acquaintances (13 per cent), home (13 per cent), and personal experience (7 per cent). Study of the data has led both Keeny and Watson to believe that isolated dramatic incidents had been given undue emphasis in the students' own narratives. Sometimes the incident reinforced an opinion already intrenched. In fact, Watson's closer analysis  of the data relating to one hundred student replies regarding attitudes toward war (187 instances of influences) shows that at least 20 per cent of the reasons given appear to be rationalizations or arguments found to justify attitudes which were already entertained. Nearly three-quarters of these attitudes seem to date back to the pre-adolescent period.
Closely similar to racial attitudes in some respects are attitudes toward other nations. One of the most enterprising large-scale investigations of international attitudes is Neumann's inquiry, using seven-point scales with 1110 high school students in the east
( 647) -ern United States. The specificity rather than generality of attitudes is strikingly shown in the fact that when questions are grouped together because of their common subject matter (militarism, public opinion, international coöperation), individual questions show mean scores varying from exceedingly favorable to exceeding unfavorable within a given field. In studying militarism Neumann makes use, for example, of such propositions as "The highest type of patriotism demands that we insist that our country never go to war again under any circumstances," and "Military preparedness is one of the best ways to prevent war." Neumann does not tell what predictive value one answer has for other answers, but his group shows strong "nationalism" in relation to the superiority of the United States over the rest of the world, but strong "internationalism" in relation to obligations of the United States toward the rest of the world.
Parallel to Neumann's study is Frederick's true-false test  given to over a thousand students in seven high schools (scattered through the country, except for the South). Over half of the students regarded it true that "as a nation, we have never mistreated the people of any country, race or nation," while 62 per cent agree that the Japanese are "treacherous and deceitful." Neumann's elaborate phraseology and careful seven-point scale, and perhaps the study of pupils in eastern high schools, seem to have missed the straight-from-the-shoulder nationalism which Frederick and others have reported. The extensive testing done by the Lynds in a mid-western city (see page 353ff.) found that of 600 children marking a true-false test, over 80 per cent agreed that "the United States is unquestionably the best country in the world," and about 70 per cent agree that "the Allied Governments in the World War were fighting for a wholly righteous cause." It ought to be remarked in passing that the Lynds find, as most investigators have found, that girls incline more toward "nationalistic" responses than boys; but the differences are too small and the causes too obscure to merit discussion.
ECONOMIC AND POLITICAL ATTITUDES
Most of the studies of economic and political attitudes will be considered below in relation to "omnibus" tests of attitude. How-ever, several, dealing solely with these issues, may be described here.
Attitudes toward occupations are one of the most interesting criteria of attitude toward the economic order. The consistency of various groups of American students in their ranking of various occupations with respect to their usefulness is very striking. When a large number of occupations were ranked by students in accordance with the "contribution to society," students of agriculture, engineering, textiles, science and business showed almost perfect agreement  In the same way, comparison of rankings by college classes showed almost perfect correlations from one to another. These data from North Carolina State College have been confirmed by almost identical data from other American institutions in other parts of the country.
In striking contrast with this is Davis's report  on the attitudes of children in the government schools of Russia. Bankers and lawyers are rated as contributing very little; hand labor, such as that of the ditch digger, is, on the contrary, rated at the top. In contrast, however, with what this group of Russian children feel about the dignity of manual labor is the all too obviously human fact that they want to get away from drudgery as much as everyone else does. Arkhangelskii  studied the attitudes of over a thousand children between nine and sixteen, the children of farmers, manual workers, officials, etc. The enormous majority regard farming and housework as unattractive, and many more want to do intellectual work than there is at present room for. The report shows the implications of this situation in connection with the desperate effort to devise an educational system which will make people want to do the kind of things which, at least for the present, have to be done.
A very successful invasion of the field of political attitudes is L. D. White's investigation  of the prestige value of public employment in Chicago. By means of house-to-house interviews in various parts of the city, more than 4000 individuals were questioned. They were asked to check, in each pair, the occupation for which they had "the higher esteem," such as (I) Stenographer, Treasurer's Office, Equitable Insurance Company; and Stenographer, Treasurer's Office, City of Chicago; (2) Janitor, First National Bank; and janitor, City Hall Building; (3) Electrical Engineer, Western Electric Company; and Electrical Engineer, Department of Electricity, City of Chicago. The answers were coded, (1) Yes, (2) "Doubtful, no difference, or unable to answer," (3) No—the Yes and No referring to presence or absence of greater esteem for city employment. In each case the subjects were told to consider each pair on the merits of the job concerned (assuming equality of salary and other factors aside from the central question of employment by the city or by a private organization). A second schedule of questions, with 690 subjects, had to do with attitudes toward city employees: e.g., "Do you get more courteous attention in dealing with city employees than in dealing with employees of private corporations ?" and "Are city employees more competent than employees of private corporations?" This schedule contains an ingenious application of the association-test technique, a series of words referring to the municipality being presented one at a time, and the subject being asked to give the first word that comes to his mind.
Much elaborate statistical material is presented. The most interesting feature is the working out of a "prestige index" which represents the difference between the sum of the prestige items for city employment minus the sum of the prestige items for private employment. The results are remarkably clear-cut. The highest of such indices are among the young (age group 15 to 19); they fall steadily as age increases. They are also highest among unskilled laborers and the lowest economic groups (as
( 650) gauged by rental value of homes), and steadily decline as economic status rises. Again, the prestige value of city employment is high among the foreign-born, and decreases with the length of time in the United States. There is again a striking negative relation between education and the prestige index of city employment. Most of these variables are obviously intercorrelated; separate items are not "partialled out." The author's purpose is simply to show the consistent effect exerted by education and improved economic status upon the idea of working for the city. The study shows what sort of esteem city employment enjoys, in order to ascertain the conditions of morale under which municipal workers carry on their duties.
This is brought out perhaps most vividly of all in the association test. To the stimulus, "City Hall," the following responses, for example, are given: booze, bums, bunk, center of graft, con-fusion, corrupt, crook, dirty, dishonest, fraud, etc. To the stimulus, "Civil Service," are given such replies as: abuses, bad, bunk, camouflage, dishonesty. Some favorable and some neutral associations are also recorded. The results here can scarcely be put upon a quantitative basis, since so much would depend on the particular word stimuli. These qualitative data, however, throw some light on the quantitative data mentioned just above.
The prestige index noted above can be compared with the series of questions in Schedule B, having to do with a series of ratings of city employees as compared with corporation employees on a number of nine-point scales, comparing the city personnel with the personnel in private corporations in respect to efficiency, honesty, courtesy, etc. The results here run parallel to the prestige results; that is, the city employees not only come off badly, but they come off worse as the economic and social status of the persons interrogated rises. Curiously enough, when the subjects are asked as to their own personal experience with city employees, it turns out that they have in general no personal grievances. In fact, reports on personal contacts with city employees seem to be favorable. It would appear, then, that condi-
( 651) -tions of chaos and confusion in the municipal government have engendered a sort of general impression or stereotype which is sufficiently powerful to counteract occasional pleasant personal experiences.
RELIGIOUS AND ETHICAL ATTITUDES
Among many people religious attitudes may not be as immediately related to obvious forms of social behavior as are attitudes on international or racial questions. Ethical attitudes, which may of course be intimately related to attitudes on international and racial matters, may also differ widely from the practices of the individuals expressing those attitudes. In spite of these discrepancies, religious and ethical attitudes are undoubtedly an indication of the way the wind blows. Attitudes may sometimes express what the individual would like to do and sometimes what he thinks he ought to do, although in both cases he may act inconsistently with the attitudes. His attitudes may be competing with his conception of the socially approved course of action. The latter, however, changes as attitudes change, and the added weight given to each attitude by the increasing prevalence of it may at any time alter the balance in such a way as to result in overt action. This suggests, of course, simply one hypothesis regarding ways in which attitudes and behavior are interrelated. This problem of the relation between attitudes and behavior has received special attention in connection with religious and ethical attitudes, because old threads in the religious and ethical fabric of life are being replaced by new ones, and the attempt to measure psychological elements in such social change is a particularly appealing enterprise.
A typical study is Shuttleworth's work with 60 sophomore students selected at random. Six pairs of contrasting terms regarding religious training were presented, the individual being
( 652) required to indicate between these two extremes (on a seven-point scale) the point where he believed his own home to lie. These six pairs gave a reliability (corrected) of +.92. The scores were compared with students' attitudes on about a hundred religious and social matters, in which, again, they decided on positions intermediate in each case between two extremes. The correlation between the amount of religious training and the degree of present religious interest and belief is rather low, but consistently positive, mostly in the neighborhood of +.2; whereas the correlation between training and present religious practice is reliably higher, +.4. The relation between present friendliness toward religion and the amount of emphasis upon religion in the home is just about zero. Here, probably, negativism aroused in some cases balances favorable responses engendered in others.
Brogan has for years been studying the question of the uniformity of ethical judgments as found among American students. Year after year he has asked his students, both men and women, to rank in order of "worseness" sixteen "worst actions" which have been over and over again listed by his classes. Year by year the group judgments of students seem to agree closely with those of previous classes, and men agree closely with women; despite the inclusion of sex offenses, the rank correlations for the sixteen offenses, comparing men's and women's scores, is +.98 +.01. In order to determine whether this unanimity of judgment was a function of the standards of the region (Texas), Brogan used the same device with a group of students at the University of Wisconsin. The correlations in various comparisons between Texas and Wisconsin judgments run from +.93 upwards.
Thurstone's study of the opinions of students with regard to the seriousness of various crimes presents a detailed quantitative check on such work as this. In fact, Thurstone presents not only the average degree of seriousness assigned to each of twenty offenses by 200 students (page 621), but shows the internal consistency or degree of variability from one student to another. It
( 653) will be clear by reference to Thurstone's table that the variability from one individual to another is a good deal greater than that obtained by comparing the average of one group with that of the average of another group. This does not mean an inconsistency between Brogan's and Thurstone's results. It is simply another case of the point which was brought out in connection with Gordon's studies (page 506) that correlations of average judgments from large groups of individuals selected from a fairly homogeneous population are bound to be greater and greater as the size of the groups increases. This is, in fact, in the realm of morals exactly what Gordon found in her study of aesthetic preferences.
The order of seriousness of transgressions reported by Brogan and by Thurstone does not differ greatly from that obtained from professional men and women, from housewives, from manual laborers, and from industrial workers by R. E. White  His list is more elaborate from the legal viewpoint than are the others. Within each of these groups, as well as in his student group, sex differences in estimation of seriousness of crimes are but slight. Some slight differences between the group averages appear. The housewives rate the sex crimes slightly higher than the other groups do. It is rather remarkable that the lists thus worked out show almost no clear differences from one group to another. White proceeds to compare the actual opinions of his subjects regarding crimes with the legal provisions regarding the severity of punishment assigned by five States, one in New England, one in the Middle Atlantic group, one in the South, one in the Middle West, and one in the Far West. These States differ from one an-other far more than do the five occupational groups (most of whose members are recruited from New England and Middle Atlantic States). Interesting questions are suggested as to the relative importance of geographical and socio-economic factors in the standardization of ethical and value judgments.
The problem of changes in standards has been approached ingeniously by Anderson and Dvorak. Thirty college men and
( 654) thirty college women participated in a multiple-choice test of attitudes; and at social gatherings of their parents and grand-parents and their friends, the tests were taken by twenty-one fathers, twenty-seven mothers, thirteen grandfathers, twenty-one grandmothers (and also by thirteen university professors and seven social workers). Fifteen behavior situations were defined, the subject being asked to state not what he would do but what standard he believes would govern his conduct in such a situation. The four criteria are (I) right as opposed to wrong, (2) prudence or intelligent judgment, (3) public opinion, (4) aesthetic standards. After a number of concrete problems had been presented in this way, the questionnaire concludes, "Upon which of the following criteria do you base your habitual conduct: (I) absolute right and wrong, (2) sense of fineness, culture and good taste, (3) intelligent and practical actions, (4) accepted public opinion and practice ?" The general reliance of the grandparents upon absolute right and wrong and the tendency of the college students to give relatively greater weight to prudence and intelligent judgment are shown in the table of average total scores for the fifteen questions:
|Right and Wrong||Aesthetic Standards||Prudence||Public Opinion|
The reliabilities of the differences are worked out by the formula D/PEo. Translated into terms of the probability that we have a true difference, we have the following figures: The chances in 100 that the difference between grandparents and students in preference for the right versus wrong standard is genuine are about 100; between parents and students, 98; between grandparents and parents, 93. Similarly, the probabilities of reliable dif-
( 655) -ferences in emphasis upon aesthetic elements are 92, 86 and 79 for the three comparisons just made. For "prudence" the reliabilities of the differences are, in the same terms, 95, 83, and 97. There does, then, seem to be for this group a technique and a definite array of results regarding changing standards.
But the authors are wary. They point out that this may not be due to a general change in standards, but to a mere age difference. The attitudes of grandparents fifty years ago, or of the college students fifty years hence, must be known, before the data could be cited as convincing proof that social standards in general, or even standards for this group, are changing. The method impresses us, nevertheless, as particularly suggestive in relation to the many devices for measuring changes in attitudes over a long period.
INNATE FACTORS IN ATTITUDES
In attempting to define the bedrock upon which social attitudes are built, it is natural to assume some fundamental disposition upon which specific attitudes inevitably rest. Thus G. W. Allport pleads for the recognition of innate patterns or groups of dispositions which may reveal themselves in specific kinds of personality, on the one hand, or in specific groups of attitudes, on the other. One cannot, in fact, understand his approach to attitudes without considering sympathetically why he rejects the conditioned-response explanation of social habits. Despite the powerful influence of the environment, he regards personality as something deeper than habit, which expresses itself through the social milieu. It is easy to see, then, why radicalism and conservatism, ascendance and submission, extroversion and intro-version, seem to him to be basic expressions of personality, not fragments somehow attached to the growing organism. We have mentioned G. W. Allport because his writings are so explicit
( 656) and so easily obtainable; he is, as a matter of fact, spokesman for a countless number whose views are just as dependent upon type theories of personality, although their language sometimes makes the fact obscure.
Two ingenious attempts have been made to discover the innate basis for attitudes, one in the field of religion, the other in that of social opinions considered collectively. Starbuck and Husband report a most ingenious effort to get at the constitutional differences between those espousing different philosophical systems. "The trouble started fifteen years ago in a Kant seminar." A third of the members of the seminar felt that Kant's arguments regarding time and space were unconvincing, another third de-fended the philosopher, while a third remained on the fence. Detailed inquiry showed that the defenders of Kant visualized time and space, while his opponents thought of them in terms of other modes of schematization resting upon organic and kinæsthetic imagery. Philosophical beliefs, in other words, seemed largely to depend upon the individual constitution. To be sure, no one knows to what extent the tendency to visualize is innate, but there is at least considerable reason to believe that innate factors play a large part in determining the capacity. Starbuck and Husband report that further investigations along this line are in progress, and that a search is being made to find objective differences in test performances between the pro- and the anti-Kantians.
H. T. Moore made an exceedingly bold and stimulating effort to discover what innate factors lie behind radicalism and conservatism as determined by Yes-No responses to twenty questions, such as the following: "Do you favor government ownership of railroads ?" "Do you believe that the advantages of trade unionism outweigh its disadvantages ?" The first five questions dealt with international relations, the next five with domes-tic politics, the next five with industrial problems, and the last five with a miscellaneous assortment. The original inquiry started from the discovery that the Dartmouth La Follette sup-
( 657) -porters in 1924 appeared to average much higher in intelligence-test performance than the Coolidge and Davis supporters; when, however, these twenty questions were given to 200 undergraduate students (Yale, Dartmouth, and Columbia), the expected superiority of those making large radical scores over those making large conservative scores did not appear. The reason for the apparent contradiction appears clearly from Moore's analysis, question by question, of the I.Q. 's of those entertaining the Yes and No answers. The intelligence-test scores of the "radicals" are high on one question, low on another; intelligence apparently plays a part in determining certain attitudes, but does not play a consistent part in driving a person toward radicalism or conservatism, as defined by Moore's test. How much genuine radicalism and conservatism there was among the groups is a question for debate; but it is of interest that when Moore asked his subjects to classify themselves on an eight-point scale from one extreme to another, 85 per cent put themselves in the two middle positions, that is, they called themselves either liberal-conservatives or conservative-liberals.
Obviously, for intensive and detailed work, it was necessary to pick out extremes. For this purpose those who gave fourteen or more radical answers were called radicals, while those who gave seven or less were called conservatives. The median I.Q. of the two groups thus constituted turned out equal. Emotional stability, as measured by the Woodworth inventory, showed no difference. Moore's analysis then proceeded to the following: Susceptibility to majority influence was tested by the technique already described (page 512). The radical group accepted 18 out of 147 chances to change their minds after hearing the majority opinion; the conservatives accepted 34 out of 109 opportunities. It occurred to Moore that the radicals might be more capable of breaking their long-established habits, and for this purpose he utilized the mirror-drawing test, in which, of course, a new habit is achieved after much wear and tear in the breaking of the old. Here the results showed that the conservatives found it
( 658) harder to learn. Perhaps, too, he thought, the radicals might tend to be quicker, and on reaction-time experiments this seemed to be true. Again, it seemed that radicals might be those who sacrificed accuracy for speed. The Münsterberg card-sorting test showed a considerable superiority in the radical group in respect to speed, without corresponding superiority in accuracy. Finally, it might be thought that radicals are people who think, in unusual terms. The Kent-Rosanoff test of word association seemed to demonstrate this. On every one of these points, then, Moore came out as conjecture would suggest.
Unfortunately for this whole approach, the data signally failed to receive confirmation when similar experiments were undertaken by Washburn and her collaborators  at Vassar. There were some cases where the percentage of radicals who surpassed the average conservative on a test was not so very different from that reported by Moore, but it appears from the data that the distributions are excessively skewed in some cases, and a considerable majority of each group surpasses the average of the other (it is, of course, possible for far more than half of any group to be above the average of the group). In other words, these intriguing results are not as conclusive as they seem.
This material from Vassar is in some respects not entirely comparable with Moore's. Extreme conservatives as defined by Moore did not seem to be very numerous at Vassar, and a slightly different method of classification had to be used in order to get the requisite numbers for reliable results. It may be, of course, that on these questions a reliable sex difference appears among con-temporary college people. It may also be that the temperamental make-up which would go with a moderately "conservative" set of attitudes at Yale or Dartmouth might go with a moderately "liberal" set of attitudes at Vassar; a person who is a middle-of-the-road conformist might be more or less radical after exposure to one environment and more or less conservative after exposure to the other, without having undergone any real temperamental change. These are all conjectures, however, and about all that is
( 659) to be said now is that neither the method of Starbuck and Husband nor that of Moore has as yet demonstrated any clear evidence of constitutional differences between those who entertain sharply contrasted attitudes.
It is, in fact, easy to speak of radicalism and conservatism, nationalism and internationalism, pacifism and militarism, without defining our terms, and then to proceed to an elaborate psychological analysis of what it is that makes for these "types." The evidence, however, in favor of the existence of such types must perforce be derived from measurements of these tendencies as they appear in different situations. A radical type, for example, would be radical not only on one issue but on many; a typical reactionary would be reactionary in enough different social situations to permit prediction from one to another.
G. W. Allport has suggested, as was seen, that attitudes are clusters or dispositions so interrelated as to be something more than disparate fragments of personality. He has, in fact, computed radicalism scores which lead him to assert that radicalism and conservatism are general traits. But he publishes only a small portion of the material upon which he bases these conclusions; he gives no intercorrelations between radicalism scores in different fields. He presents no tabulation of scores on the various issues which would make it possible to tell whether the building up of a large radical or conservative score is anything more than the occasional product of independent causal factors, in the same way that an occasional throw of six 6's at dice might happen despite the lack of causal connection between the scores of the separate dice. Vetter,37in turn, has asserted that there is such a thing as general radicalism, basing this statement upon the fact that some individuals do in fact yield large radicalism scores. The crucial test, however, is whether a radicalism score on one scale gives prediction as to the score on another scale, and data on this point are not supplied. Vetter's contention that he has established the existence of such a general trait as radicalism because of the high reliability between the halves of the
( 660) test will not bear analysis. He states that the correlation between scores on eighteen scales and scores on the remaining eighteen scales is as high as + .70. He spoils his point, however, by explaining that the same number of religious questions, the same number of sex questions, the same number of pacifism questions, etc., appear in the two halves. Correlations between two halves, each of which is made up of a large number of elements, might, of course, approach or even equal r.00, if the weighting of all the elements entering into one half were the same as that entering into the other. A person's radicalism on one economic issue might be balanced against his radicalism on a closely related issue appearing in the other half of the test, and so on.
Symonds's evidence  is against the interrelation of different types of radicalism and our own (as yet unpublished) studies are in the same direction. To look for the general characteristics of "a radical"—for example, to ask what his general intelligence or his emotional stability is—seems therefore to us to be probably a wild goose chase. It might, in fact, prove valuable to know the average intelligence quotients of radicals on some particular issue or group of issues; but until the interrelation of the different kinds of attitudes has been definitely proved, this search for personality traits leading back to the supposed types seems to us full of logical difficulties.
The only studies of which we know that seem to throw light directly on this problem of types (interrelations) of social attitudes are those of Lundberg  and of George. Lundberg used as subjects a group of over zoo students, who took a questionnaire comprising four sets of twenty questions each. One set of twenty had to do with religion, another with politics, another with economic issues, and another with domestic and moral issues. He defines as a "variant" a person who is in a minority of 15 per cent or less on a given question, that is, a person who is outvoted on a given question by an 85 to 15 majority. On this basis it turns out that 63 per cent of all the subjects are variants on one issue or another; in other words, most of them tend to
( 661) pick one or more unpopular opinions. The question now arises as to whether there are such persons as chronic variants, that is, those who tend in general to disagree with the majority. The data are emphatically in the negative. There are only two individuals who are variants on as many as four issues out of the eighty, and three who are variants on three issues. It will be recognized at once that if the tendency to be a variant on one question is unrelated to the tendency to be a variant on another, there would, among such a large number of subjects, have to be certain cases showing more than one deviation. It is rather surprising that out of over two hundred individuals not a single one should be found who had more than four out of eighty such attitudes. This is not a direct answer to our question about types, but it will, we think, be admitted to be at least a challenge to the type theory. The results may perhaps depend on this special definition of a "variant."
Some evidence is found in R. W. George's study (cf. page 663) of the interrelationship between liberalism and conservatism on national and international issues. George used a questionnaire consisting of forty items of the Yes-No type, the Yes or the No answer having been previously defined as liberal or conservative on the basis of the judgment of a group of graduate students of economics. With a group of over one hundred subjects, George found a correlation of + .55 +.05, between liberalism on domestic issues and liberalism on international issues. This is, as far as we have discovered, the most direct evidence for the conception of a general disposition to some sort of social attitude.
The following facts must, however, be remembered: that the students were in a metropolitan environment subject to an unusually large amount of "liberal" or "radical" literature; in fact, that it is practically impossible for such students to avoid con-tact, direct or indirect, with internationalist literature. Such internationalist literature—The Nation, for example, or The New Republic, or a book by men like John Dewey or Bertrand Russell—inevitably contains references to domestic issues as well.
( 662) The correlation for these given subjects may mean nothing more than that they are subjected to certain opinions on international issues at the same time and in the same way that they are subjected to certain opinions on domestic issues. This, of course, does not explain why the journals or the books tend to present certain interlocked sets of attitudes or opinions, and we should, of course, recognize that there remains a plausible case for the doctrine of a psychological interlocking of opinions. But our experimental evidence is the best we have, and, as this illustration will point out, it is far from clear that the results obtained for the group of metropolitan students can be regarded as meaningful. It must, of course, be remembered in any event that the findings may as likely be sociological as psychological; that is, they may be due to common elements or clusters of opinion-making forces in society rather than to such inherent psychological determiners as are frequently expressed in the notion of "the maladjusted radical," the "smug conservative," etc.
That there are clusters of personal characteristics we have many times had occasion to point out and to emphasize (cf. pages 204 and 581). The problem discussed here is simply the problem of clusters of opinions on public issues.
It is interesting to note how the characteristics of the "radical" and the "conservative" differ as definitions and methods of investigation and interpretation differ. One discovers, for example, the following interesting antithesis: "In personality rating studies a suggestive negative correlation has been found between the trait of radicalism and the possession of insight. That is, there was a tendency for those who were judged as extremely radical to be judged low in insight, and vice versa." This was written by F. H. Allport. and a footnote follows which reads thus: "The writer is indebted for this information to the researches, not yet published, of Dr. G. W. Allport.” Five years later G. W. Allport  published data pointing to the following conclusions: Radicals possess greater insight than do conservatives; they are less prejudiced and better scholars. The study of their char-
( 663) -acteristics leads one, G. W. Allport says, to react favorably to Shaw's statement that any man under thirty who knows the facts about the social order and is not a revolutionist is an inferior. Out of shop there may be places where the exaltation or defamation of radicals or conservatives may be legitimate. For the present we are inclined to believe that data so confusing as these fail to carry us very far; we do not even know exactly who is being exalted or defamed.
R. W. George undertook to determine the emotional differences between liberals and conservatives by means of the Pressey X-O test. It will be recalled that this test presents many series of words, the subject being asked to cross out in one series all the words which are unpleasant, in another series all the words which describe things which he thinks are wrong, and in an-other series all the things about which he has ever worried. He then goes through the various series again, drawing a circle around that word in each five which designates the thing which is the most unpleasant, the most wrong, or the thing about which he has most worried. A person with worries like those of the people about him tends, in respect to worries, to be "normal"; thus the important thing is to study the number of cases in which he worries about something which is not typical of his group. A large number of unusual responses would give a high "idiosyncrasy" score.
It seemed that liberals might perhaps be those who had some-what more unusual emotional reactions than conservatives. A preliminary inquiry containing some items from Moore's test, described above (page 656), was first given to about fifty students in a graduate course in economic theory, who were asked to indicate in each case whether the Yes or the No answer was the "liberal" one, and to mark each item with an interrogation point if uncertain. Almost complete agreement existed with regard to a few of the questions to be classified, and forty out of
( 664) the original sixty were agreed upon to an extent shown by 75 per cent assigning a given answer to the liberal column, the doubtful and the dissenters being outvoted three to one or more heavily (the doubtful votes ran much higher than the dissenting votes, the actual disagreement on these forty questions being slight). These forty questions were now submitted to three classes (one in economics, one in social psychology, one in tests and measurements) comprising, in all, 104 students. The Pressey X-O scores obtained from these same subjects correlated + .55 + .04 with the liberalism scores. In view of the smallness of the probable error of this correlation, it seemed surely to be worth while to repeat these methods with undergraduate classes and we have twice done so. In both cases the correlations, though positive, have been below + .30, and have therefore suggested that George's procedure is simply an interesting lead, well worth following, to be sure, if the method can be improved, but scarcely capable of throwing much light upon the psychology of liberalism. An important source of error in group tests of radicalism' and conservatism has, moreover, been shown by Vetter;44individuals taking a test in the group are much more conservative than when taking the same tests alone (compare Allport, page 516).
In order to show the complexity of the effort to find any clear-cut relation between the items which are supposed to play a part in the determination of attitudes, we offer this table of correlations between Army Alpha (intelligence test), Watson fair-mindedness test, Pressey X-O idiosyncrasy, and R. W. George liberalism. This is based upon forty-eight cases in an under-graduate psychology class; it does not differ strikingly from a similar inquiry made the succeeding year.
|Alpha and Pressey||– .22 ± .09|
|Alpha and Watson||– .15 ± .09|
|Alpha and George||+ .36 ± .08|
|Pressey and Watson||– .03 ± .09|
|Pressey and George||+ .13 ± .09|
|Watson and George||+ .19 ± .09|
The highest of these correlations is between intelligence and liberalism; this is very similar to Vetter's and G. W. Allport's results. But it is too low to mean much. All the concepts lying back of these tests (including that of intelligence) require far more analysis.
One more search for the emotional predispositions to types or groups of social attitudes deserves to be mentioned.
Among the psychoanalytic approaches to this problem, one of the most interesting is Klein's study of the influence of attitude toward one's father. In seventy questionnaires filled out by men students, some questions had to do with the attitude of affection or fear at different periods in life, and some were concerned directly with memories as to ways in which the individual tried to adapt to paternal discipline. One question reads, "Did you plan revenge phantasies to get even? (a) no revenge planned, (b) a few, (c) frequently, (d) a systematic series of revenge phantasies." All the data were studied by three judges, who assigned to each individual a certain degree of "father antagonism." A questionnaire on public issues was then given, consisting primarily of opportunities to reveal extreme radical answers (not simply liberal or progressive). The correlation co-efficient between radicalism and father antagonism thus turns out + .6o. With seventy cases, this ought to be fairly dependable, but knowing the vagaries of statistics, it would be better to suspend judgment until it is reported with another group.
We are inclined, then, to believe that emotional sources of sets of attitudes have been hard to establish. But another possibility for explanation in terms of "emotionality" remains open; this is the explanation of specific attitudes in terms of specific causes.
PSYCHOLOGICAL EXPLANATIONS OF PARTICULAR ATTITUDES
An interesting analytical study of what it is that makes a person "typical" or "atypical" in a social attitude is Allport and
( 666) Hartman's work on student attitudes on a number of public issues. A large number of spontaneous opinions having been written on such questions as prohibition, the League of Nations, and the distribution of wealth, six judges (members of the faculty) classified the opinions and made up scales ranging from one extreme to the other on each issue. These scales were then given to a large number of subjects who were asked to indicate where they stood on each scale (e.g., the degree of "wetness" or "dryness" with which they sympathized, the positions ranging all the way from the "open saloon" to the rigorous enforcement position). The same individuals were also asked to state the degree of their confidence in each opinion.
The results showed that in general there was a piling up of opinions near the middle of the opinion scales, while extreme positions tended to be less numerous. On the other hand, the confidence scores showed in several cases a peculiar tendency. Individuals at the extremes were actually more confident of their positions than were those at the middle. This led Allport and Hartman to the conclusion that the individual who finds him-self in the minority must find some device for justifying to him-self the opinion he expresses, and that he resorts to the emotional device of suffusing his opinion with the emotional character of a conviction. This peculiar relation between confidence and "atypicality" does not appear in all cases, but this is a valuable and suggestive hypothesis. Personal interviews with the students are said to have revealed in some cases emotional factors connected with these replies.
The very elaborate data of Vetter  have analyzed somewhat more intensively this question of atypicality, as well as the question of "radicalism." In addition to 36 opinion scales, Vetter made use of intelligence tests, the Colgate C2 self-rating test for extroversion-introversion, the Allport ascendance-submission test, and various other devices. Family income, order of birth, and political party preference are among the variables considered. Vetter's data do seem to indicate that scores on the psycho-
( 667) neurotic inventor} are slightly higher among atypicals than among typicals, but certain statistical considerations, particularly in relation to striking sex differences, make it doubtful whether the result is significant. The radicals, considered as a separate group, do not appear to differ in any significant way from others in scores of this sort. In a recent (unpublished) study, Vetter has tried to find what it is that causes a certain opinion to be characteristically classified by judges as a radical or a conservative opinion, etc. After intercomparison of "radical," "liberal," "conservative" and "reactionary" opinions, he came to the conclusion that three criteria had been responsible for the judges' decision as to how, to classify an opinion. These three he calls "moralistic attitude versus pagan," "individualistic attitude versus paternalistic," and "favoring trying the new versus op-posed to trying the new." It would seem, then, that the concepts of "radical" and "conservative" are at least capable of being broken down (in this set of questions) into three more or less separate variables. Further analytic study of this sort is, of course, badly needed.
The low correlations so often found between the amount of information and tire degree of enthusiasm expressed in any given social attitude have suggested, as in the work of Lund (page 683) and of Allport and Hartman, that certainty of opinion is a function not of information but of desire. Only one direct study, as far as we know, has been carried through on the problem of the relation between information and certainty. This is the elaborate work of H. S. Carlson, based on students' attitudes and opinions regarding Mr. Hoover and Mr. Smith during the fall of 1928. After each answer the subject reported the degree of his certainty on a scale from o to 5 (inclusive). The information score was based on a series of very searching questions regarding the backgrounds and life histories of the candidates, and their stand on a number of campaign issues. The information score increased greatly with the class in college,
( 668) but the certainty score changed very little. On the other hand, considering 357 cases individually, a correlation of + .21 between information and certainty was discovered. That this is not a simple linear relation, however, is shown by the fact that a certainty score of o usually means low information, while a score of 1 means high information; that is, people who are especially well equipped on an issue tend to be low on certainty, but not at the bottom. On 206 cases intelligence test scores were available; these correlated + .28 with information (no partial correlations were carried out). One study is no basis for generalization, but on the basis of this study we are inclined to challenge the general tendency to assume that certainty is an easy device for covering up lack of rational justification for one's views.
A study dealing in part with the same problem of the relation of attitudes to information is Harper's inquiry regarding the social attitudes of American educators. He constructed scales upon which responses were to be classified as (I) extremely conservative, (2) conservative, (3) liberal, (4) radical, and (5) extremely radical, and undertook to find out, by a nation-wide sampling of educators, what distribution of opinions was held on a large number of public issues. Typical Yes-No problems are: "(1)In teaching the vital problems of citizenship teachers should so impress on the students the approved opinions in these matters that life's later experiences can never unsettle or modify the opinions given; (2) If our people were willing to try the experiment fairly the government ownership of railroads would be for the best interests of the country." A large number of judges were asked to decide whether the Yes answer to each question was to be classified as tending toward the conservative or the non-conservative end of the above five-point scale, and very high agreement was found. In all, 2900 educators were tested, every State in the Union being represented. "Conservatism" seemed to be related to lack of experience with the prob-
( 669) -lems in hand. Such lack of experience resulted in great inconsistency from one question to another. The relation was striking, the scores in the left-hand column indicating the degree of "liberalism" or "non-conservatism:"
|Those Making a Liberalism Score of:||Average Number of Inconsistencies|
Another study throwing light on the relation of attitudes to knowledge is
of "World Citizenship."
He submitted a large number of questions of fact, and some questions of opinion
as well, to a group of 43 "competent judges," recruited especially from social
scientists, religious and moral leaders, and persons deeply concerned with
international relations. Manry believed that the answers given by these judges
to a large number of questions relating to international relations could be used
as a standard against which to compare the "world citizenship" of college
students. Having ascertained in each case how a group of competent judges
thought or felt about a point, it was possible to say to what extent a college
student was alive to the considerations which prompted such an opinion. Examples
of Manry's questions are: "A vigorous nation (a) tends to expand territorially
but may meet its growing needs in other ways, (b) must expand its geographical
area, (c) must expand if it has insufficient coal and iron in its own borders."
One hundred per cent of the 43 competent judges preferred opinion (a). Of
several hundred college freshmen, 81.4 per cent likewise did so. An example of
tests on questions of fact is the following: One word out of the series
at the right of each line is to be underscored as indicating "the country which
best exemplifies" the caption at the left, for example:
Consumers' coöperation England, France, Portugal
Irrigation Egypt, Panama, Poland
A matching test is likewise used, the left-hand column giving
( 670) names of places and the right-hand column describing events which have happened at these places within recent years.
The test is reasonably reliable, the odd-even method giving an r of .73, which (by Spearman-Brown) becomes + .84. The correlation between the information items and the judgment items is + .69. Data obtained on various small groups point to a relation of r = + .40 to r = + .70 between test scores and intelligence. The most striking results of the test are the differences between colleges, and, in some cases, between underclassmen and upperclassmen in the same college. Dartmouth freshmen who had had an orientation course scored higher than those who had not. (The situation was ideal for experimental purposes, since half had the course the first semester and half the second semester, the sectioning being arbitrary.) The score of the former group is 89.6, against 78.2 for the latter (S.D. not given).
In relatively few cases, however, are causes of high scores easily named. Certain intangible factors seem consistently to favor eastern institutions and those on the Pacific Coast. Significant, how-ever, is the higher score of Jewish students at Columbia and Harvard on both the information and judgment tests (again S.D.'s are not given, but the differences seem impressive). The influence of information upon opinion might be emphasized here, and it might be possible to say that greater information as to foreign countries, public happenings, etc., is a cause of consistent differences of opinion. We have, however, no measure of the opinions of those who teach the students, and no measure of the degree to which the students are positively led by their teachers (or, on the other hand, antagonized by them to take the opposite position). Lacking such information, it is quite possible that those teachers who teach certain items of information also lead their students to adopt certain kinds of opinions. As far as the students are concerned, then, it would not be possible to prove that their own information affected their own opinions, and we should have to make a statistical study of the teachers.
METHODS OF PRODUCING CHANGES IN ATTITUDES
Until a few year s ago, one might well have said that no precise and quantitative data on the subject of the effects of propaganda and publicity were in existence. Several good studies are now available.
The purpose of Carr's investigation  was to ascertain the effectiveness of anti—tuberculosis publicity in a number of counties in Michigan. Granting that it would be desirable to have trust-worthy and comparable data on the relation of publicity to actual tuberculosis mortality, Carr confines himself to (I) the actual publicity techniques employed in these counties, and (2) the results in terms of knowledge about tuberculosis and related subjects. Data were first assembled on "population per cent urban"; population changes; per capita bank deposits, 1926; average value per farm; and many other variables. Two counties, Kalamazoo and Saginaw, were selected for detailed comparison. The data selected. as bearing upon the efficiency of publicity, comprised the following items: seal sale, use of clinics, use of hospital headquarters, and early diagnosis projects. Both counties were found. to have used press, movies, schools, letters, circulars, special tagging, exhibits, billboards, professional groups, and personal contacts. Each organization had films displayed in three motion picture theaters; neither county made much use of billboard service which had been donated. Activities through these various channels are classified as "wholly adequate," "partially adequate," "no opinion," "partially inadequate," and "wholly inadequate," on the basis of judgments from physicians, organization workers, the investigator, and laymen.
The two counties were then compared in terms of the answers given by 480 persons in one, and 340 in the other, to such questions as the following: (I) Which one of these three diseases killed the most people in the United States last year: cancer, tuberculosis, heart disease? (2) Cancer may affect almost any part
( 672) of the body. In this respect, it is like: pneumonia, appendicitis, tuberculosis. (3) If you thought you were getting tuberculosis, the thing to do would be to (a) exercise in the fresh air every day, (b) keep on as you are and refuse to worry, (c) get some medicine from the drug store, (d) go to a good doctor at once. The differences in "percentage wrong" are in general small, but they show remarkable consistency. On "the general nature of tuberculosis," 42.2 per cent failed in Kalamazoo, 46.8 per cent in Saginaw; on "three things to do to protect yourself if someone in the house has tuberculosis," 44.8 per cent failed in Kalamazoo, 51.8 per cent in Saginaw. Not the absolute size, but the consistency of these differences through these tests, is important, and this is exactly what would be expected from the fact that the Kalamazoo organization was pretty consistently ahead in the utilization of publicity methods. Discussion of Carr's results by G. J. Drolet  makes the whole situation clear by pointing out the fact that in Kalamazoo there is a paid executive, while in Saginaw a group of persons do as well as they can without the services of a paid officer.
Both counties show clearly that propaganda is strikingly unsuccessful at certain points and strikingly successful at others. About 99 per cent in Kalamazoo and 93 per cent in Saginaw fail on the question: "Name one thing that the local tuberculosis association has done this year beyond asking for money," while 55 per cent in both counties know (can draw) the symbol of the Tuberculosis Association. Even with small budgets and in competition with hundreds of other sorts of publicity, the Associations have impressed on both communities a scattered array of useful items of information. Carr's technique for the study of publicity is valuable in checking up on the better results which seem to have been obtained where the superior organization would lead us to expect it.
More ambitious and much more important for our purpose is Biddle's study  of propaganda and counter-propaganda. After giving an elaborate opinion test, and some reading-matter con-
( 673) -taining high-powered propaganda, Biddle had six groups of students (350 subjects in all) read a series of pamphlets explaining the technique of propaganda. These pamphlets, though they made rather hard reading for some of the high school students, gave concrete and detailed explanations of propaganda devices. Typical materials on war propaganda were, for example, presented in full, and the resulting effects on the various warring populations explained. At the end of the study period (which averaged three weeks), the subjects again took an attitudes test. As far as they knew, this concluded the experiment. Two weeks later, however, they were again given the violent propaganda material. Now, whereas the material of the pamphlets to be studied dealt with domestic issues and the "Atlantic Relations" of the United States (that is, with our contacts with Europe, the Caribbean, etc.), the propaganda citations just mentioned dealt exclusively with the "Pacific Relations" of the United States, with emphasis on China, Japan, and the Philippine Islands. There was, therefore, no actual overlapping of content.
In each of the six groups the nine lessons were used, the teachers administering the material and the tests. The schools varied widely, four being public and one private, and their composition (socio-economic and "racial") varying considerably. One of the six is a college group. For every experimental group there is a control group at the same educational level (11th or 12th grade, or freshman year of college) undergoing class instruction in some topic identical or closely allied with the one under discussion in the experimental classes. For example, a 12th-grade economics class is a control for another 12th-grade economics class in the experimental group.
The tests for susceptibility to propaganda show rather high reliability, the odd items correlating + .77 +.03 with the even items, which (by Spearman-Brown) comes to + .87. On the other hand, responses appear to be highly specific in relation to each national group toward which attitudes are tested. A strong dislike for the Chinese may be accompanied by enthusiasm for
( 674) the Filipinos. An anti-Japanese but pro-Chinese attitude was found rather commonly (it is worth recalling that the students were all on the Atlantic Coast).
The criterion of susceptibility to propaganda was the opinion of twenty-four judges thoroughly familiar with the issues, but themselves differing widely in opinion. "Susceptibility to propaganda" is therefore not simply a stigma attached to someone who happens to disagree with a given opinion; it is rather the yielding to an emotional argument which the bulk of a group of experts regard as unreasonable. Of the twenty-four judges, seven were university professors, seven were members of the Institute of Pacific Relations (American group), and five were graduate students of political science. The scoring key was made up by choosing the mean of the judges' markings. Three questions on which the judges scattered widely were thrown out. In some cases skewed curves or other irregularities appeared in the ratings of the items by the judges (the scale would probably have done its work better if such cases had also been thrown out, and if only those materials had been included upon which practically unequivocal judgment was rendered by persons differing sharply in personal opinions).
The test of "gullibility" (i.e., the extreme statements on Pacific Relations) was "announced as a study of opinion. No connection with the teaching material was made. . . . Any students who connected the gullibility test and the teaching material themselves made the transfer between the two" (italics ours).
Despite a few cases of difficulty in understanding some of the material on the part of some of the high school students, the results are extremely clear-cut, the experimental reading matter having an exceedingly pronounced effect. It is especially to be noticed that the great decrease in susceptibility to propaganda, does not relate merely to the increasing resistance to fire-eating, jingoistic material. On the contrary, both nationalist and inter-nationalist propaganda (irrational appeals having as their purpose the exaltation of American interests or the exaltation of
( 675) interests of other national groups) are taken more sagely. The scores are as follows:
The controls, it will be noted, made a slight gain, but very much less than that of the experimental group. The actual difference is so huge (17.5 times the standard deviation of the difference in the one case, and 37.1 times the standard deviation of the difference in the other) as to be as near absolute reliability as anyone could wish; in fact, the differences are several times as big as is necessary to make them "reliable." Individual differences in initial gullibility are great. In general, those with highest gullibility gain most from the experimental material (r =+ .48).
Ninety-six students from all experimental groups were studied for an explanation of their gain (or failure to gain) in resistance to propaganda. "Improvers" and "non-improvers" are compared. The improvers, as was seen before, are more gullible at the outset than those who remain the same, but they scatter over a considerably wider range of initial scores. Intelligence plays a part in improvement, as is shown by the fact that between percentiles 6o and 100 on intelligence appear 58 per cent of the improvers and only 46 per cent of the non-improvers.
When the subjects were asked to write what they thought of the propaganda material, it was found that the improvers specifically showed that they understood they were being subjected to propaganda in twice the proportion shown by the non-improvers (27 against 14). ("When the difference 27—14 = 13 is reduced to a proportion of 48, it becomes .56—.29 = .27. The standard deviation of this difference is .014, making the difference again significant.") As was noted above, whatever transfer was made
( 676) between the lessons and the final test was largely made by the subject's own recognition of the relevance of the teaching material. This is borne out by the fact that improvement is so definitely related to the conscious recognition of the fact that propaganda is being used. Similarly, most of the cases where there is no improvement are due to failure to get the point of the study material with relation to the test.
To discover the transfer value of printed to other verbal material, Biddle went into one of the schools with a propagandist speech (the pupils did not know him by sight). He delivered a vigorous bit of nationalist propaganda with a good deal of bitterness and distortion of economic facts abroad. The students, including both experimentals and controls, were asked to write comments on the speech. Twenty-four of the pupils disagreed with the speaker; eighteen of them came from the experimental group. Twelve agreed enthusiastically with the speaker; of these, nine were members of the experimental group. But of these nine, seven were members of the experimental group who had failed to improve as a result of the teaching. Tentatively, it would seem that the result had carried over even to the different situation.
Does the order in which arguments are presented play any important part in the determination of attitude? Lund  drew up elaborate and cogent paragraphs of argument for and against each of three general propositions, e.g., "Should all men have equal political rights ?" Six groups of students, twenty in each group, were asked to read this material, the affirmative arguments being read first and the negative arguments last by three of the groups, and the reverse procedure employed for the other three groups. The subjects rated each proposition on a "belief scale" from –10 to + 10 before reading, after the first reading, and after the second reading. The results are much more striking than one would have guessed. For group A, for example, the average rating on the belief scale for the first proposition is only
( 677) + .5. After reading the affirmative arguments the average rating jumps to + 6.8. After reading the negative arguments, it falls to + 3.3. Taking the data as a whole, it is clear that the material which is first read has a great deal more influence than that which is read later. Lund therefore postulates a "law of primacy in persuasion."
Another device used to measure changes in attitude is the scale of certainty administered to test the influence of public speakers. Some years ago, Rice and Willey undertook to measure the influence of W. J. Bryan's address on evolution by asking students who had heard the address to indicate their attitude toward the evolutionary theory, together with the degree of confidence which they had in their own judgment, and to give the same data for their opinion and the degree of confidence before hearing the address. It is interesting to note, first, that the address produced a very considerable number of changes in degree of certainty, many students, for example, who had been definitely evolutionists becoming somewhat weakened in their conviction, while a few reacted antagonistically, that is, became more hostile to Mr. Bryan's view than they had been before.
There is, of course, no sharp line between "propaganda" and "education." The influence of education on attitudes has been but very inadequately studied. The influence of the college course is itself almost entirely unknown, some authors confidently asserting consistent changes from freshman to senior year, and others doubting not only the consistency but the rôle played by the college in such changes as do occur. We might well consider a few attempts to determine what the college course does to our attitudes. Twenty-five short declarative sentences were shown by E. S. Jones to about 250 freshmen, about 75 upper classmen and nearly 100 second-year law men. These propositions were to be marked from + 2 (positive agreement) to -2 (positive disagreement). Five of the statements were related to national and
( 678) social optimism ("Modern advertising is a social blessing"), five related to labor and economic problems ("Poor men cannot get justice in the courts"), five related to discipline ("One cannot be healthy and at the same time smoke tobacco and drink strong coffee"), etc. The replies which Jones considers "reactionary" are as common among seniors as among freshmen, although a slight trend toward greater economic conservatism and greater religious radicalism seems to be suggested.
Symonds  has made an attempt to get a picture of the entire educational field from the eighth grade to the senior college class in Hawaii. From this might be derived a picture of the effect of education upon attitudes. Over 100 questions dealing with all sorts of contemporary problems were used. In each case five judges were asked to indicate their opinion as to the "liberalism," "progressivism" or "radicalism" of an answer (these terms being used as equivalent words in opposition to conservatism). If the judges voted four to one or five to zero that a Yes answer was liberal, it was so considered. Trends toward liberalism seemed to appear with increasing age in connection with certain questions, especially those relating to government control and government benevolence. Amount of information is, however, correlated + .28 with liberalism, and it is hard to say what intrinsic changes are actually going on here beyond the sheer recognition of what the questions mean. Since Symonds does not publish his entire questionnaire, it is quite possible that the replies which had been classified as conservative were in general merely stupid or uninformed replies. This would load the dice and produce a spurious correlation.
At Stanford University a questionnaire on current political, religious, and ethical questions was printed by the college daily, and the results were worked up by Willoughby. Since only 800, that is, a quarter of the student body, replied, all results must be taken with great caution on account of sampling errors. In general, Willoughby's data are, however, similar to those obtained under other conditions. Notably, he finds a "trend toward
( 679) liberalism with increasing exposure to academic influences," except on a few issues. Many sex differences appear. The "trend toward liberalism" seems, from his data, to be an evidence that college does make changes either in attitudes on the questions or in the increasing willingness of certain persons to hand in questionnaire replies.
How much change in opinions on social issues results from the sociology course itself appears not only to be an obscure problem, but to have led as yet to the invention of no really reliable techniques. Several studies, such as those of Young, have reported that practically no change in attitudes has been produced either by rational or emotional appeals, or by a combination of the two; others, such as those of Zeleny, indicate definite and striking changes. Young's indirect method (asking the questions in such a form that the student is not directly reminded of the fact that his agreement with the instructor is being gauged) seems generally preferable. On the other hand, the real test would seem to be the administration of an elaborate set of in-formation and attitude tests by strangers who are known to be working up the anonymous data for their own purposes and in no way in collusion with the instructor. Attempts in this direction are only beginning.
The whole machinery of propaganda and education can be better understood in the light of the theory of "stereotypes." Our attitudes toward races, nations, flags, national anthems, and to-ward the words which crystallize generally accepted values, such as freedom, honor, and democracy, tend in general to be imprinted upon us in a more or less standardized form; it is for this reason that Lippmann  has described them as "stereotypes." When we hear of patriots or bolsheviks, we can scarcely stop to
( 680) consider the individual attributes of all the individual persons thus designated. A stereotype of a patriot or of a bolshevik, loosely regarded as serving to describe all members of the species, also serves the purpose of all but the most sophisticated and careful forms of thought. Where no thought at all, but only emotion, is involved, the stereotype is an infinite labor-saver.
Rice  experimentally analyzed the function of stereotypes in relation to nine photographs taken from the Boston Herald. These represented Herriot (then French Premier), Duncan (vice-president of the American Federation of Labor), Krassin (Soviet ambassador to France), Agel (a bootlegger), Schwab (steel magnate), Heinz (of the 57 Varieties), Pepper (United States Senator), and others. These photographs were shown to nearly 150 students, together with a list of terms such as "premier," "labor leader," "bolshevik," "financier." Individuals had to pick out the right word for each man. The stereotyping tendency to agree that a certain kind of picture goes with a certain appellation is perfectly clear-cut, not only where the designation is right but where it widely misses the mark. There is, in other words, a tendency to a stereotyped conception of what a labor leader or a financier looks like, and this, even in the case of mistaken identifications, produces a converging of judgments. In the case of Krassin, a winged collar, a Vandyke beard, and a mustache led to "fifty-nine identifications as the United States Senator, in comparison with nine as a bolshevik and none as a labor leader." A similar experiment was tried on twenty-five members of a farmers' grange with very similar results; in fact, the rank correlation between farmers' and students' judgments was + .84.
After this experiment was terminated, three groups of students graded the nine individuals on the two traits, intelligence and craftiness, the latter being defined as "that characteristic, the possession of which would lead to the taking of an unfair advantage in a business negotiation."The first group of students did not at the time know the identity of the men whose pictures were shown. The second group were given false identifications based
(681) upon the stereotyper already discovered. The third group was informed of the real identities. The ratings on intelligence and craftiness are in the direction expected. Disclosure of the true identities led to changes of rating in the same direction among both students and grange members in the case of seven out of the nine individuals rated.
Stereotypes have been examined also by Sargent. His concern is chiefly with the standardized emotional response which members of large social groups display when confronted with certain words. Accordingly, Sargent emphasizes emotional response to words, many of which probably lack the characteristic "pictures in our heads" to which Lippmann refers. The study has to do not with pictures, but with immediate affective responses.
After measuring the affective reaction of a group of persons to 50 words, he undertook to ascertain to what extent their stereotyped responses- could be altered by reflection. To what extent can the emotional tone of the stereotype be changed when the actual meaning of the concept is scrutinized? A list of common words relating to social, economic, political, religious, scientific, educational and recreational activities was read aloud, one at a time, to a group of students, and each member of the group reported his immediate emotional reaction (within 3 seconds) on a five-point scale (2, very favorable, 1, moderately favorable, 0, indifferent or don't know, –1, moderately unfavorable, –2, intensely unfavorable. The first ten words from the list are: prosperity, Chicago, skating, socialist, self-denial, Monroe Doctrine, prostitution, philosopher, war, Greta Garbo; and the last ten are: gravitation, Mussolini, mysticism, pickpocket, Holly-wood, bravery, beggar, idealism, pessimist, foreigner. Thinking of a stereotype as a word which tends to produce positive or negative reactions (with a relatively small number of neutral reactions), he proceeds to this definition: "Those words are considered stereotypes where +2 and +1 reactions total more than the sum of 0, – 1 and –2 reactions, or whose – 2 and – 1 reactions total more than +2, +1 and 0." The five words
( 682) producing the most favorable stereotypes for a small group of graduate students at Columbia University are:
|The sum of +2 and
+ 1 reactions
|0||The sum of –1 and
To measure the reliability of such a test is difficult, because a second test made immediately after the first would be greatly influenced by memory of the previous responses, while too long a time interval would permit many real changes in attitudes. The time interval selected was four weeks.
For thirty-three subjects participating in both experiments, the average number of changes was 11.5 in a positive direction, and 12 in a negative direction. Immediately after the second taking of the test, each subject was instructed: "In the space beside each word on these sheets, give the meaning or image each term has for you, or possibly a short definition or synonym. After so treating a word, indicate your considered emotional reaction to it, on the five-point scale to the right."
The most definite finding of the experiment is the small amount of change in the emotional reactions. After such reflection, the average change is about two-thirds as much as that produced by the four-weeks' interval. As Sargent points out, one can scarcely be expected to forget an opinion just recorded. It is natural, if only in self-defence, to cling to an opinion or even a mood just expressed. Only three words ("pacifism," "sex," and "democracy") showed a greater change in emotional coloring after reasoning than they did as a mere result of the time interval. The greatest change in opinion after both the second and
( 683) third tests occurred in those words which were least stereotyped for the group as measured by the first test. This is, at least in part, a result of the scoring method, since the + 2 and -2 re-actions had no chance to fluctuate in the direction away from the zero point. Taking the results for what they may be worth, they point to a rather small amount of change in stereotyped emotional response which may be expected to occur as a result of brief reflection.
Such studies suggest an intimate relation between thinking and feeling, or, if one prefers to emphasize the conative aspect of life, greater emphasis on wishing than on thinking. This has been directly examined by Lund, in a rather elaborate statistical analysis of the relation existing between belief on the one hand and certain intellectual and emotional factors on the other. He constructed a scale for belief (see page 676) ranging from + 10 (belief allowing for no doubt) through the point of absolute uncertainty, down to –10 (disbelief allowing for no doubt). A certainty scale was constructed ranging from + 10 (very certain that you can affirm it) to –10 (very certain that you can deny it). In the same way a desire scale ranged from + 10 (highly desirable) to –10 (highly undesirable).
Thirty propositions were presented which had to be rated on each of these scales. For example, "Is a democracy the best form of government?" "Does death end personal existence?" "Do two plus two equal four?" After preliminary experiments, another measuring device was added, requiring the subject after each proposition to indicate to what degree the proposition was based on "indisputable evidence," and to what degree it was a "matter of opinion," measured in per cent; the sum of the two percentages would be 100. The relation of belief to desire was studied with groups of students at Columbia College, Barnard College, and the University of Nebraska. Taking the belief and desire co-efficients for the groups (243 students in all), the correlation between belief and desire was + .81. The correlations between belief and evidence are much lower; not all the results are given,
( 684) but apparently the results given by a group of thirty Columbia sophomores are considered typical, the relation being expressed by a correlation of + .42 +.10. Anticipating the comment that this dependence of belief upon desire (more than upon evidence) is a function of the special questions used, Lund calls attention to the fact that, using Sumner's older set of questions on belief,  thirty-one students showed a correlation of + .76 between belief and desire, and + .69 between belief and evidence.
"Is desire," Lund asks, "an important antecedent of belief, or is it merely a consequent ?" He concludes that the comparatively low correlations between belief, on the one hand, and knowledge and evidence ratings on the other, point to the causal efficacy of desire in relation to belief.
Up to this time we have been speaking of the degree to which propositions are believed in, or desired, in terms of the pooled responses of groups. Obviously we need to go farther and ascertain the coefficients of correlation between belief and desire when each individual's reactions are considered. These are mostly between + .55 and + .85; they are not strikingly different from the group figures.
If, then, desire and belief are so intimately connected, what is the subject's attitude toward the determination of his own belief? Thirty-five students were asked, in relation to each belief rating, "What led you to rate it as you did ?" Out of about a thousand responses received, the following classes of causes are constructed: Teaching and training, 317; personal experience, 142; personal opinion, 116; personal reasoning, 92; desire and satisfyingness, 54; authoritative opinion, 46; public opinion, 44; axiomatic principle, 28. The subjects were then asked to arrange these determinants in their order of merit, the degree to which opinion might legitimately be determined by each factor. The following sequence appeared:
1. Personal reasoning
2. Axiomatic principle
3. Personal experience
4. Authoritative opinion
5. Teaching and training
6. Public opinion
7. Desire and satisfyingness
Surely, they did not know what swayed their opinions. When Columbia and Nebraska students were asked to rate the relative importance of the factors which influenced the "average college student," and when these were compared against the estimates of the factors influencing their own opinions, it appeared that in general the subjects regarded their fellows as more swayed by teaching and training than they themselves were, but themselves as more influenced by personal reasoning and personal experience.
Unfortunately we know very little about the influence of the press upon attitudes. In comparison with the material upon what the press is publishing and what is being read, there is singularly little definite information as to the influence of the press.
The experimental technique used by Bird  in studying the influence of the press on accuracy of report is, however, a good beginning. He seized upon an opportunity presented by the fact that a college newspaper had distorted one of his lectures and had thus subjected many of his students to an influence which could be directly compared with what he had actually said. His lecture told, among other things, how college students had, in Seashore's experiment on suggestion, reacted to illusory warmth 415 times out of 420 cases. Another point made in the lecture was that in the experiment on hand rigidity carried out by Aveling and Hargreaves, 49 per cent of the children experienced this effect of suggestion. The university newspaper carried an article on suggestibility occupying a quarter column in the lower left part of the front page. This article stated that "98 per cent of college students are so suggestible that if you tell
( 686) them their hand is rigid they believe it," and similar gross distortions of the lecture.
All university students received the newspaper in their mail boxes. Some, however, did not read the article in question. The lecturer prepared four questions to determine the extent to which the students had been influenced by the article: (1) "Who were subjects in an experiment on hand-rigidity produced by verbal suggestion?" (2) "What percentage of these subjects were suggestible?" (3) "How many college students were used in the experiment on the illusion of warmth ?" The fourth question asked whether the student had read the college newspaper article on suggestibility. Three groups of students participated, a group of men preparing for business, a group of men students in the College of Science, Literature and the Arts, and a group of women in the same. Of the business group, 85 had read the article, 70 had not. Of the other group of men students, 75 had read it, 33 not; the figures for the women are 166 and 58, respectively.
The result is definite. In the case of every question with every group, the percentage of correct answers is higher among those who had not read the press article, while the percentage who gave the "press" answer is higher in those who read the article than in those who did not. For example, three-fourths of the business group who had not read the paper gave the lecture answer to question (I); one-fifth gave the press answer; a few gave nothing. (By accident, of course, a certain number of those who had not read the paper might give the press answer; similarly, a few might happen to "guess" the lecture answer though not remembering it.) Bird concludes, "In answer to all questions pertaining to the lecture material the groups of students who have not read the press are more accurate reporters than those who have read the press. . . . Of the eighteen differences .. . six are approximately three times the probable error of the difference and six are approximately four times the probable
( 687) error." Since the quizzes were given on different days of the week, it was possible to study the possible influence of the lapse of time, but no clear difference appeared. There was also no evidence that outside discussion of the quiz influenced the answers.
END-RESULTS OF FORCES WHICH SHAPE ATTITUDES
The end-result of all these processes which have just been described might be expected to be a good deal of highly irrational belief existing among us all, whether we have been subjected to the process of education or not. And this is exactly what the various studies of superstition have found. Beliefs which almost everyone in sophisticated circles derides, such as the unluckiness of 13 or the luckiness of a four-leaf clover, are nevertheless admitted as genuine beliefs by 15 or 20 per cent of college students, and other equally irrational beliefs not quite so "naïve," such as chose having to do with signs and portents (black cat brings bad luck), are reported by considerably larger numbers. To stretch time term "superstition" a little, and to include under the head of popular "misconceptions" positive belief in phrenology, palmistry and the like, would give plenty of data on the very general acceptance of propositions and practices which betray a general love of the marvelous or, perhaps, even a desire to be mystified. We should like, if we could, to trace this matter of superstitiousness to its origins in the innate constitution of people. The statistical study of college freshmen by M. E. Wagner  indicates that suggestibility, as measured by the Aussage test (page 168), gives a correlation of + .25 with amount of superstition (92 individuals). The figure is "reliable," but is not high enough to throw much light on the matter. There is a very low positive correlation between age and the amount of superstition which has been .dropped by the individual in the course of his life. The relation between superstitiousness and intelligence is
( 688) negligible (in this group). For the present, then, it would be best to consider superstition as sufficiently explained by individual backgrounds. Wagner suggests that superstitiousness is accept-able in women in many cases where it would be laughed at in men, and points to the fact that the group of men have dropped about half of the superstitions they ever held, whereas the women still cling to about five-sixths of theirs. This kind of superstition, however, seems to be rather "lightly worn." Even a semester's course in psychology is reported by Gilliland  to have knocked many of these beliefs flat.
That we carry around with us plenty of superstitions—political, economic, ethical, as well as religious—common observation, as well as the above data on propaganda, will testify; and courses in psychology usually do not knock them flat. We live in a world of irrationalities; and from the point of view of the year 3000 we shall probably be seen to have been swimming in a sea of superstition not so profoundly different from that of the time of Attila.
Looking back through the last two chapters, it will be clear that the concern with individual differences is no special perversity of the psychologist, but, rather, the natural expression of a demand which the very nature of the subject matter has imposed. If men were like molecules, we could hope for a science of social psychology based on the same sweeping generalizations, the same scorn for personal peculiarities which the physical sciences show. In the very midst of every psychological generalization however, comes the question of individual difference; and in social psychology, more so than in the psychology of the isolated individual. The laws of social psychology are, in fact, in part laws as to the relative influences of causes—laws which pertain to the very nature of the individual differences which appear. If biology, psychology, and sociology achieve such exactness that individual differences among men can themselves be stated in
( 689) terms of universally valid laws, so that the reason for every difference is in terms of a general principle, we shall breathe a great sigh of relief at the simplification which our science will undergo. Until such laws have been established, it will be well to regard every problem in human social conduct as in part a problem in "the laws of human nature," and in part a problem in analysis of the peculiar qualities of individual persons.
But here, as is so often the case, we have not succeeded in experimentally analyzing the things we should most like to know about. Shall we, then, retire to the study, and think it out? We are earnestly told to do so. But the whole history of science and human affairs cries a warning. In concluding, we must emphasize this warning, and try to appraise what the experimental method can justly claim. The human mind has proved again and again to be too meager and too fragile an instrument to dig, unaided, deep into the flint-like structure of reality; it can suggest hypotheses, but only pick and shovel can find whether the predicted discovery of truth will occur. In some of the chapters which have preceded, dozens upon dozens of ingenious hypotheses have had to be packed off, protesting, always, as they go; but in each generation of scientists there are many who are incapable of learning what the history of science and of philosophy shows—that the "thinking through" method of probing reality is one which is denied to the sons of men. Thinking through to the point of testing—yes, this is of the very essence of science; but thinking through to the point of deciding—this is for the gods, and for them only.
To fumble around with pick and shovel before one has thought through to the point of testing is just as idle and wasteful a process, and it is one nearly as frequently seen. This volume has had many occasions to point out that some of the data are meaningless because a clear question was not asked of Nature in the first place. In relation to the endless question whether facts or
( 690) theories come first, we can only say, on behalf of social psychology, that the problem is like that of deciding which foot shall be put first in walking. There are those who wish to walk entirely on the theory foot; others, on the fact foot. Science has always walked like a true biped—a step of theory, a step of fact, a new theory, a new fact—each hypothesis leading to patient search for evidence, and each new fact leading to extension of theory. But hopping on either foot is better than standing still in despair lest we never learn to walk. The goal and the path are exciting; and light upon both is beginning to appear.