Some Needed Concepts in the Study of Social Attitudes
Muzafer Sherif 
PERHAPS more effort and words have been poured into the study of social attitudes than into any other single problem area in social psychology. Under the labels of attitude and attitude change or one of the mushrooming neologisms, research on social attitudes continues to increase. This is as it should be.
From childhood on, the individual encounters objects, persons, and groups with insistent labels of approval, disapproval, or other value shadings attached to them by people important in his eyes. Of course, the process is seldom a one-way street. In the interaction process between him and others, the individual's desires as they are formed at the time, his strivings to belong and to prove himself play their part.
Whether his social attitudes are formed through interaction with other individuals or are primarily shaped by dictums, pronouncements, or exhortations of others, the state and particular brand of an individual's socialization can be expressed in terms of the attitudes which he has formed relative to stimuli within his psychological world. Stimuli relevant to his attitudes are selectively perceived and reacted to in a characteristic way. In no small measure, his appraisals, and hence his reactions to these stimuli are in terms of his attitudes towards them. Appropriate attitudes are involved when he is interacting in small groups, when he is exposed to communication from newspapers or television, and when he is deciding on a course of action. The importance attached to the study of attitudes is understandable.
What is not entirely understandable is the nebulous state of conceptions in this central problem area. No wonder that research results have frequently been difficult to evaluate. Until recent years, no greater confusion prevailed than in research attempting to change social attitudes through communication. Reported results were inconclusive and even contradictory. Significant changes toward the position presented in communication were reported, but so was "no change"; and still other studies resulted in change in the opposite direction (Murphy, Murphy, & Newcomb, 1937; Williams, 1947; Hovland, 1951, 1954).
Much confusion may be avoided in research if the prop-
(196) -erties of social attitudes which have made their study essential in any scheme of social psychology are kept in clear focus. Attitude change studies gain coherence and predictions become possible when the properties of the individual's attitude are analyzed in relation to the properties of the stimulus material to which he was exposed.
Within the limits of available space, concepts found useful in recent research on social attitudes will be discussed, a study analyzing latitudes of acceptance and rejection on a political issue will be summarized, and the implications of this mode of analysis for attitude change studies will be indicated.
PROPERTIES OF SOCIAL ATTITUDES
Social attitudes are learned relative to a stimulus or stimulus class, such as social objects, other persons, groups, or institutions. Thus they always imply a subject-object relationship (Thomas & Znaniecki, 1918) . The positive or negative values which the individual attaches to things, persons, or groups are usually outcomes of interaction between him and others, in which process his desires to be accepted, to belong, to prove himself, to amount to something play a crucial part. Thus a major source of the content of social attitudes is the values or norms of the individual's reference groups.
What does it mean for the individual to have an attitude towards a group, a church, a country, a political or economic issue? The attitude defines his positive or negative relatedness to its referent. Once a stimulus class is charged with value, the individual sees things related to it in a characteristic way, in a selective way, and he reacts accordingly. It is from the characteristic and selective mode of behavior in
(197) given conditions that an attitude is inferred (Campbell, 1955).
Events in line with his attitude are desired and satisfying. Events out of line or contrary to it produce dismay, annoyance, and disappointment. Thus a social attitude has the essential earmarks of a motive. There is good reason for using "sociogenic motives," "acquired drives," and "social attitudes" as equivalent terms.
The relationship defined by an attitude is not usually expressed in neutral terms, but in emotionally charged terms. A Baptist, for example, does not express his attitude toward his church by saying: "I formed a positive opinion of this particular church, and therefore I prefer it." He says: "I am a Baptist." A disrespectful remark from an outsider hurts and, if possible, a corrective reaction is meted to the offender.
Much of what is included in one's sense of self-identity and self-esteem consists of subject-object relationships implied in the social attitudes formed during one's particular life history. If not transformed under specifiable conditions, social attitudes are lasting and not transitory affairs. Confusion in definitions of social attitudes can be reduced by including their motivational and (more or less) lasting properties therein. Transitory or motivationally neutral opinions and sets produced for the performance of laboratory tasks can be designated by concepts other than "attitude."
CATEGORIZATION OF ITEMS RELEVANT TO AN ATTITUDE
What is the consequence of having a positive attitude towards one's own group? That group is placed high in acceptance and other groups are ranked in descending order, usually in terms of definitions of those groups prevailing in
(198) one's own reference group. Similarly, upholding a stand on a controversial issue amounts to categorizing other stands on the issue say, along the acceptable-objectionable dimension, in terms of their relative distance from the one the individual upholds.
The behavior from which attitudes are inferred always involves a judgmental process. In 1934, Gardner Murphy made me keenly aware of this general fact by calling my attention to an observation by Wells (1908) . Wells noted that the ordering of a series of weights was in terms of the objective gradations of the series, whereas the ordering of pieces of music was in terms of the preference standard of the individual.
Even when order remains constant, the grouping of items into categories may vary as a function of attitude. In a recent study by Carolyn Sherif (1958), subjects sorted 100 cards into as many or as few categories as they saw fit. For one group of subjects, a number from 1 to 100 appeared on each card. For the experimental group, the same numbers appeared preceded by a dollar sign, and the instructions were to consider each card as a price tag on a garment. In the first case, the cards were categorized in terms of convenient groupings of the decimal system. In the second case, categorizations clearly reflected the income and prestige levels of the subjects, and the resulting differences in what was considered "cheap," "reasonable," "dear," etc.
When the individual has a definite attitude relevant to stimulus material, he brings established categories to the task of dealing with it. His own stand delineates the bounds of his tolerance or acceptance, and is customarily a major anchor in his judgments. This set of categories constitutes his reference .scale, for the placement of items in that universe of discourse. In our research we have frequently encountered spontaneous
(199) protests from a subject that the scale he was instructed to use did not extend far enough, or that it was too extensive for him to make proper ratings. Closer examination of such cases revealed lack of fit between the categories imposed by instructions and the reference scale the individual had formed in regard to the issue at hand.
Categorization of items relevant to attitudes is equivalent to evaluation, and evaluation presumes the placement of items into categories. The judgmental and motivational processes producing the behavioral outcome are inextricably intermingled.
SOURCE OF THE INDIVIDUAL'S REFERENCE SCALE AND A BASELINE FOR RESEARCH
In the psychophysical experiment, the subject judges a stimulus as to weight, intensity, or extent relative to a standard stimulus of specified objective value, or to the repeated presentation of a series of graduate stimulus values. The keenness of his judgment, the effect of the series, and anchoring effects are gauged against known stimulus values.
In social life, the individual is ever passing judgments on a social issue in terms of the reference scale he has formed in his previous encounters with that issue. The reference scales used by different individuals vary in range and in the widths of the segments each finds acceptable and objectionable. Here the scale of categories prevailing in the individual's reference groups, ranging from the limits of acceptable categories to the limits of unacceptable categories on the given issue, is a useful counterpart of the objective series values against which judgments are gauged in psychophysical research. The categories prevailing in a group provide a psychological scale against which the differing ranges of individual
(200) reference scales, the different stands upheld by different individuals (including the deviates or nonconformists) can be assessed.
The limits of possible positions on an issue are the bounds of conceptual categories available in the social setting. For example, one would have difficulty if he asked the Siriono of Bolivia to group more than 3 items into categories because their number system consists of 1, 2, 3, and then "much" or "many." Likewise, in a social setting where wines are not differentiated and labeled by different names, it would be foolhardy to ask subjects to rank wines as to, say, their appropriateness for different occasions. In a setting where there are no different shadings on a political issue, the available categories would accommodate only similarity to the prevailing stand and "different."
The use of forced choice between two stimuli at a time (paired comparisons) and establishing a rank order on this basis are, of course, always procedurally feasible. But the results of this method reflect the discrimination ability of the person rather than the cutting points which characteristically define the direction, tolerance limits, and rejections of the attitude in question. The primary and crucial cutting points in attitude research are the limits of categories defined as acceptable and as unacceptable. From the point of view of attitudinal categorization, say, on racial discrimination, the fundamental step is to ascertain that being discriminated against in housing, in work, or in school are all placed in an undesirable category by those subjected to them. Ordering within the acceptable and unacceptable categories is a technical step that should follow. But if investigation starts with forced choices, the main boundary conditions are obscured, and the cutting points arrived at in the analysis may be sheerly arbitrary. Beebe-Center (1932) made a similar com-
(201) -ment on the limitations of forced choice in discussing judgments of pleasantness and unpleasantness.
In sum, the prevailing psychosocial scales categorizing positions represented in a sociocultural setting and prevailing in an individual's reference groups can be used as a baseline in research. Against this baseline, the relative positions of particular individuals, the widths of their tolerance range, the degree of their conformity or nonconformity can be advantageously computed.
CHARACTERISTIC DISTRIBUTION OF JUDGMENTS AND NUMBER OF CATEGORIES AS AN INDEX OF AN ATTITUDE
By focusing upon the judgmental processes involved when the individual reacts to stimuli relevant to an attitude, certain predictions were possible (Hovland & Sherif, 1952; Sherif & Hovland, 1953) . The predictions were based (1) on the demonstrated effects of affectively charged internal anchors on the placement of items in a series (e.g., Volkmann, 1936; Hunt & Volkmann, 1937); and (2) on the demonstrated liability of ambiguous stimuli to displacement as a function of motivational factors (e.g., Chapman & Volkmann, 1939; Proshansky & Murphy, 1942; Marks, 1940). These predictions were as follows:
1. Individuals with strong attitudes (i.e., ego-involved) on an issue will tend to bunch together a disproportionately large number of items in extreme categories at the expense of intermediate categories.
2. Individuals with moderate positions and little egoinvolvement with an issue will distribute their judgments more evenly throughout the categories.
3. The items which are liable to displacement towards ex-
(202) -treme categories will be middle-of-the-road (neutral) items and not those sharply defining a position.
4. Individuals upholding strong stands will concentrate more judgments at the extreme segment of the scale opposite to their own stand on the issue. In other words, highly egoinvolved individuals upholding stands near the extremes will be "choosy" in accepting items, thus lumping a much greater number of judgments into the segment of the scale which is objectionable from their standpoint. This well-known tendency is expressed by partisans of public issues in the form: "Those who are not definitely for us are against us," and may be conceptualized as a raised threshold of acceptance and lowered threshold of rejection.
5. When individuals are given freedom of choice as to the number of categories they will use in placing items on an ego-involving issue:
a. Persons with strong attitudes will use relatively fewer categories and will concentrate their judgments at the two extremes of their "own scale," the greater frequency being at the extreme opposite to their own stand.
b. Persons with moderate positions will use a more extended scale consisting of a relatively larger number of categories than that used by those upholding an extreme position, and will distribute their judgments more evenly throughout the scale.
These predictions were tested in research on judgments of the favorableness or unfavorableness of statements to the social position of Negroes. Of the 300 subjects (college students), about one-third were Negro students. The statements sorted were 114, ranging from those clearly favorable to the social position of Negroes through definitely anti-Negro expressions originally used by Hinckley (1932) . The sortings were made under two procedures:
1. Imposed categories procedure: Ss were instructed to sort the statements into 11 categories.
2. "Own categories" procedure: Ss were to choose as few or as many categories as they saw fit, with category one being defined as "most unfavorable."
In the first session, part of the subjects judged under the imposed category condition and the others under the "own categories" procedure. In the second session about two weeks later, the sorting conditions for the two groups were reversed.
The results support the predictions stated above. The checks introduced make explanation of the results in terms of carelessness in sorting extremely unlikely. Subjects were not told at the first session that a second session would occur. Yet those subjects who concentrated their judgments at the extreme segments of the scale under the imposed category condition were the ones who used only three or four categories under the "own categories" procedure. The same pattern of frequency appeared under the "own categories" condition in more accentuated form, with a disproportionately large number piled in the extreme category opposite to the subject's own stand on the issue.
It is suggested that the "own categories" procedure for judgment in which a sufficient number of middle-of-the-road or ambiguous items are presented may be used as a quantitative "projective" technique in assessing attitudes on controversial issues. The subject is instructed to sort the items in terms of their "pro" and "con" nature. The number of categories used and the degree of concentration of judgments in extreme categories reveal the strength of the attitude. The characteristic distribution of judgments reveals the position that the subject upholds, the neutral items being displaced predominantly toward the. extreme opposite to his own stand on the issue. Perhaps the "own categories" procedure may be
(204) useful in assessing the rigidity or flexibility of the person in given matters.
LATITUDE OF ACCEPTANCE AND LATITUDE OF REJECTION
We have seen that a person strongly committed to a stand on an issue becomes highly selective in admitting statements to categories acceptable to him. His threshold of acceptance is raised, and his threshold for rejection is lowered. Thus he tends to lump together a disproportionately large number of items into objectionable categories. This tendency has been observed frequently. For example, Johnson (1955) reported that persons holding extreme stands towards war gave negative judgments more frequently and with greater confidence than did individuals with more moderate positions.
For the study of reaction to communication and attitude change, the consequence of raised and lowered thresholds can be measured in terms of latitudes of acceptance and rejection for items on an issue. The customary use of a single score or average to represent the individual's attitude on an issue obscures these significant relationships. Individuals upholding the same modal point on an issue do differ concerning their personal involvement and the other positions on the issue which they will also accept or tolerate. They do differ in regard to the limits of their tolerance, beyond which all other positions on the issue are flatly rejected as objectionable, even as obnoxious.
The latitude of acceptance may be defined as consisting of those positions on an issue that the individual finds most acceptable, plus other positions also acceptable to him. The latitude of rejection, consists of those positions on the same
(205) issue that he finds most objectionable, plus other objectionable positions on the issue.
Our first attempt towards assessing latitudes of acceptance and rejection, carried out in 1948-1949 by the writer, C. Hovland, and E. Volkart as a pilot study, utilized the "closed shop" issue. The second was carried out in 1952-1954, and utilized the prohibition issue in Oklahoma (Hovland, Harvey, & Sherif,1957). The unpublished research to be summarized here was carried out in Oklahoma as part of a collaborative research project with Hovland, and utilized the issue of the 1956 presidential election.
On the basis of the previous studies, the hypotheses were:
1. The latitude of rejection of individuals committed to an extreme stand on an ego-involving issue is greater than the latitude of rejection of individuals with a moderate position on the same issue.
2. The latitude of rejection of individuals committed to an extreme stand on an ego-involving issue will be relatively greater than their latitude of acceptance.
The study was carried out in two sessions of which the first was devoted to procedures designed to ascertain the attitudes of the subjects in terms of their latitudes of acceptance and rejection. Here we can only mention in passing the results of the second session in which communication was presented and reactions to it obtained.
The issue of the presidential and vice-presidential candidates in the 1956 election was suitable for testing the hypotheses because adequate numbers of subjects upholding differing stands were readily available and because the issue held nation-wide interest at the time.
For the purpose of securing latitudes of acceptance and rejection on the election issue, nine statements were prepared after extensive pretesting. The statements ranged from an extremely pro-Republican position through a middle-of-the-road position to an extremely pro-Democratic position. For purposes of analysis, the nine positions were designated by the letters A through I. Statements A, E, and 1 are reproduced below:
A. The election of the Republican presidential and vice-presidential candidates in November is absolutely essential from all angles in the country's best interests.
E. From the point of view of the country's interests, it is hard to decide whether it is preferable to vote for presidential and vice-presidential candidates of the Republican Party or the Democratic Party in November.
I. The election of the Democratic presidential and vice-presidential candidates in November is absolutely essential from all angles in the country's best interests.
The complete set of nine statements (A to I) was mimeographed on each of four sheets of paper. On the first sheet, subjects indicated the statement "most acceptable" to them; on the second, they indicated other statements also acceptable to them. The positions checked on these two sheets provided the data on their latitudes of acceptance. On the third sheet, subjects checked the statement "most objectionable" to them, and on the last sheet they checked other objectionable statements. The checkings on these last two pages yielded data on their latitudes of rejection. Subjects were not forced to check any statements in regard to which they chose to remain noncommittal, and such results were classified accordingly.
The statements were designed to prevent a "ceiling effect." fn fact, our results reveal a reluctance to endorse extreme
(207) end items (statements A and 1) , even by subjects selecting a position next to the extreme as the one most acceptable and actively committed to one of the major parties as campaign workers.
Results are based on data from 406 subjects from college populations in three universities in the Southwest. The median age was slightly over 21 years, and 78 percent of the subjects were 25 years or under. A special point was made to obtain members of the Young Republicans and the League of Young Democrats who were engaged in campaign activities at the time (the month preceding the election). In addition, some small groups in university dormitories with known stands on the issue were obtained. The responses of these subjects served to check the validity of the results.
In part of the analysis, subjects were classified according to the position they chose as "most acceptable." Then the frequencies with which these subjects accepted, rejected, or remained noncommittal on each of the remaining eight positions were determined. From this analysis, nine graphs were constructed, each representing the frequencies of responses by subjects with a different stand on the election issue.
On each graph, the various positions on the issue are shown on the abscissa, ranging from A (the extreme pro-Republican position) to 1 (the extreme pro-Democratic position). The ordinate represents the percentage of subjects upholding a given position who accepted, rejected, or remained noncommittal to the other eight positions on the issue. The first graph, for example, represents the acceptances, rejections, and noncommittal responses of subjects checking the most pro-Republican position (position A) as "most acceptable." The adjacent graph represents the responses of subjects checking the most pro-Democratic position (position I) as "most acceptable" The solid line
|Latitudes of acceptance and rejection for individuals upholding given positions on a controversial issue.|
represents the percentage of subjects accepting given positions. A double check mark is located above the "most acceptable" position, which is, of course, 100 percent in each figure. The long-dash line represents the percentages of subjects rejecting given positions, and the short-dash line indicates the percentages of noncommittal responses. The most objectionable position for subjects is represented by XX. For example, the most objectionable position for subjects at position A is the I position.
A definite criterion was necessary in order to make generalizations about latitudes of acceptance and rejection of individuals who find different positions as "most acceptable." The criterion used was the placement of a given statement within the "acceptable," "rejected," or "noncommittal" category, as the case might be, by 50 percent or more of the subjects. In each figure a heavy dash line represents this 50 percent cutting point. When a percentage of acceptance or rejection is above the 50 percent line, we included that posi-
(210) -tion in the latitude of acceptance or the latitude of rejection of that subject group.
Note that the distributions for subjects holding pro-Republican positions (graphs on the left for positions A-C in particular) are the reverse of those for subjects holding pro-Democratic positions (graphs on the right for positions 1-G) . The figures are presented side by side for easy comparison, and it may be seen that one of each pair appears to be almost a mirror image of the other. The figure for subjects at the middle position (E) is at the bottom. These subjects include equal numbers of positions in the latitude of acceptance and latitude of rejection. The distributions for rejection and the noncommittal classification are bimodal.
The results shown by these graphs quite clearly support our hypotheses, namely, that the latitude of rejection of individuals upholding extreme stands is greater than the latitude of rejection of individuals upholding moderate positions, and that, for individuals with extreme stands, the latitude of rejection is greater than the latitude of acceptance.
We were also interested in determining whether our hypotheses would be supported by analysis of the relative sizes of the latitudes of acceptance and rejection for each individual. Therefore we divided the subjects into three classifications on the basis of their own positions on the issue: extreme pro-Republicans (checking positions A, B, or C as most acceptable); extreme pro-Democrats (at positions G, H, or I) ; and moderates (at D, E, or F). In each of these classifications we determined the number of individuals whose latitude of acceptance was greater than their latitude of rejection, the number whose latitude of acceptance was smaller than their latitude of rejection, and the number whose latitudes of acceptance and rejection were equal in size. All
(211) of the differences in the resulting 3 x 3 table were in the predicted directions.
The overwhelming majority of subjects upholding extreme positions had latitudes of acceptance which were smaller than their latitudes of rejection, while this pronounced trend was not evident in the case of moderate subjects. Chi square analysis yielded a value of 26.34, which is significant at less than .001 level for 4 degrees of freedom. Separate analysis indicated significant differences between extreme pro-Republican subjects and moderate subjects, and between extreme pro-Democratic and moderate subjects. Differences between extreme pro-Republican and extreme pro-Democratic subjects were slight and not significant, as was predicted.
SOME IMPLICATIONS FOR REACTIONS TO COMMUNICATION AND ATTITUDE CHANGE
One implication of our research is that reaction to communication presenting a given position on an issue will not depend solely upon the relation between the position the individual chooses as his "own stand" and the position advocated in communication. His judgment of the communication is affected also by the distance of the position advocated from the limits of his latitude of acceptance. We found, for example, that a communication presenting a stand somewhere near the middle-of-the-road is judged differently by moderate subjects with broad latitudes of acceptance than by subjects with extreme views and narrow latitudes of acceptance (Hovland, Harvey, & Sherif, 1957) .
The individual whose latitude of acceptance extends to a position close to that presented in communication or propaganda is likely to assimilate the position presented in communication, i.e., to regard it as closer to his own stand on the
(212) issue than it actually is. As a result he is more likely to be influenced by the content of communication.
The individual with a narrow latitude of acceptance whose limit is farther from the position presented in communication reacts quite differently. In this case, the result is akin to the contrast phenomena in judgment and perception. The individual judges the communication as more divergent from his own stand on the issue than it actually is. He is irritated by the communication. He may even become more uncompromising in his stand than he was initially. We suspect that the relationships and judgment processes involved are pertinent to an explanation of the well-known boomerang effects in propaganda.
From a theoretical point of view, the results demonstrate the relevance of assimilation and contrast effects found in laboratory studies of judgment and perception to the study of effects of communication advocating a point of view on a controversial social issue.
Beebe-Center, J. G. Pleasantness and unpleasantness. New York: Van Nostrand, 1932.
Campbell, D. T. The indirect assessment of social attitudes. Psychol. Bull., 1950, 47, 15-38.
Chapman, D. W., & Volkmann, J. A social determinant of the level of aspiration. J. abnorm. soc. Psychol., 1939, 34, 225-238.
Hinckley, E. D. The influence of individual opinion on construction of an attitude scale. J. soc. Psychol., 1932, 37, 283-296.
Hovland, C. I. Changes in attitude through communication. J. abnorm. soc. Psychol., 1951, 46, 424-437.
Hovland, C. I. Effects of the mass media of communication. In Lindzey, G. (Ed.). Handbook of social psychology. Vol. 2. Cambridge: Addison-Wesley, 1954. Chap. 28.
Hovland, C. I., Harvey, O. J., & Sherif, M. Assimilation and contrast effects in reactions to communication and attitude change. J. abnorm. soc. Psychol., 1957, 55, 244-252.
Hovland, C. I., & Sherif, M. Judgmental phenomena and scales of attitude measurement: item displacement in Thurstone scales. J. abnorm. soc. Psychol., 1952, 47, 822-832.
Hunt, W. A., & Volkmann, J. The anchoring of an effective scale. Amer. J. Psychol., 1937, 49, 88-92.
Johnson, D. M. The psychology o f thought and judgment. New York: Harper, 1955.
Marks, E. S. Skin color judgments of Negro college students. J. abnorm. soc. Psychol., 1943, 38, 370-376.
Murphy, G., Murphy, L. B., & Newcomb, T. M. Experimental social psychology. New York: Harper, 1937.
Proshansky, H., & Murphy, G. The effects of reward and punishment on perception. J. Psychol., 1942, 13, 295-305.
Sherif, Carolyn W. Categorization of valued items as a function of the individual's own reference scale. Paper presented to the Southwestern Soc. Sci. Ass., April, 1958.
Sherif, M., & Hovland, C. I. Judgmental phenomena and scales of attitude measurement: placement of items with individual choice of number of categories. J. abnorm. soc. Psychol., 1953, 48, 135-141.
Thomas, W. I., & Znanieeki, F. The Polish peasant in Europe and America. Boston: Badger, 1918.
Volkmann, J. The anchoring of absolute scales. Psychol. Bull., 1936, 33, 742-743.
Wells, F. L. On the variability of individual judgment. In Essays philosophical and psychological in honor of William James, by his colleagues at Columbia University. New York: Longmans, Green, 1908.
Williams, R. M., Jr. The reduction of intergroup tensions: a survey of research on problems of ethnic, racial, and religious group relations. Soc. Sci. Res. Council Bull., 1947, 57.