Adolescent attitudes and behavior in their reference groups within different sociocultural settings[1]

Carolyn W. Sherif and Muzafer Sherif

IN SOCIAL psychology, the focal problem is the relation between the individual and his sociocultural setting. In this paper, we shall focus on our continuing research program initiated over ten years ago, which was conceived as basic research on the individual-group relationship during adolescence. The principal objectives of the program are theoretical and methodological.

It is our firm conviction that if basic research is to bear on the actualities of social behavior, and hence have valid applications, it must come to grips with at least these two major problems of theory and method: first, the interdisciplinary nature of the problem of the individual-group relationship, and second, the resulting necessity for attacking the problem with a combination of research methods, including both experimental and field methods, integrated in such a way that precise and valid accounts of the phenomena in question can be given.

In psychology, the unit of study is the individual; the data are his behavior during interactions that occur within an ecological and a cultural setting. These interactions and settings are the stimulus situations. Their complexity dictates that psychological data cannot be discrete bits of be-

( 98) -havior lifted from their context. On the contrary, we have to study behavioral events over time, in sequence, noting their consequences in shaping the individual's attitudes and actions toward others, and toward significant activities and objects.

A basic assumption in our research has been that human social behavior, whether directed toward socially desirable or undesirable ends, can be accounted for more adequately within a context of interpersonal relationships. Characteristically, these interpersonal relationships occur within the bounds of groups —the family, age-mate groups, school and community groups —whose acceptance or rejection, approval or disapproval, and attribution of status and prestige are of serious personal concern. This means that, in addition to analysis of behavior at the psychological level, analysis at the sociological level is needed to conceptualize the stimulus side of the individual-group relationship. We have to place the attitude and behavior of the individual within the organization and the set of values or norms that count for him.

But this is not all. Small groups and interpersonal relations within them are not closed systems. Their organization, norms, and members' activities are affected by relevant features of the settings in which they interact —the nature and spatial arrangements of housing and recreational facilities, the standard of living, the cultural and educational level of the populace, and the prevailing reference scales defining what is undesirable and what is desirable in many respects. Finally, these specific settings are part of a larger society which conveys values and goals to its members, especially through the mass media and the educational system.

What is needed is research that specifies the ecological and sociocultural settings as they actually confront the individual, that locates him within the set of interpersonal relationships which count for him, that determines the values prevailing among people around him, and that relates his behavior over time to this context. Through the interrelation of findings at these various levels of analysis, the prediction of individual behavior may begin to appear feasible. The huge task of completing such research necessarily requires that we employ a combination of research methods and techniques — not merely side by side in insulated compartments, but in a coordinated way. Then, and only then, can we truly integrate the findings from experimental research, on the one hand, and field research, on the other.

During the course of our experiments on group formation and inter-

( 99) -group relations (Sherif, 1966; Sherif, Harvey, White, Hood, & Sherif, 1961; Sherif & Sherif, 1953), we developed procedures for the observation and measurement of interaction processes. Research on natural groups of adolescents extended and greatly refined the observational procedures (Sherif & Sherif, 1964). Both the group experiments and a series of laboratory studies permitted the development of methods to secure short-cut behavioral indicators of interpersonal attitudes, status and role expectations, and intergroup attitudes through cognitive functioning — namely, judgment and perception (see Sherif, 1967). Thus, our methodology was derived partly from experimentation and partly from field research in the course of the program itself.

The focus of the research program is on natural groups of adolescents as they interact in their natural surroundings, unaware that they are subjects of research interest. In this report, we shall first consider the properties of natural groups, derived both from a survey of the research literature and our own research. Next, we consider why such groups form so frequently among adolescents in modern societies and why they are so important in psychological development, especially in the re-formation of the self system during that transitional period of growth. With this background, we shall turn to findings from our research program, first summarizing illustrative generalizations about the relations between the sociocultural setting and individual attitudes; second, indicating our methods of research with emphasis on the study of interpersonal relationships and conformity-deviation relative to group norms; and finally, citing several experiments to illustrate the coordinated use of field and laboratory methods in the effort to attain valid generalizations.

The Research Program on Natural Groups of Adolescents

What are natural groups? Why is their study important? Among other things, natural groups are not casual encounters with people; they are not classroom groups in a school; they are not ad hoc collections of N persons of the kind studied in the great bulk of laboratory studies of "small groups." Nor is a natural group something that drops out of the blue in one piece, or that forms by spontaneous generation. Natural group is a convenient term from the point of view of the initial formation of the human group. A natural group is a human formation, consisting of a specifiable number of members, formed over a time span, through voluntary and informal interaction of individuals. In time, a natural group may or may not

( 100) be codified as a recognized and formal organization. Thus, the natural group is not an esoteric phenomenon. The formation of a natural group seems to be a universal human process in various spheres of social life.

The minimum essential properties of any group consist of (a) a pattern or organization of roles and statuses and (b) a set of values or norms, at least in spheres of activity frequently engaged in by the group. One can say that the groupness of the group is directly proportional to the degree that the set of norms is shared by the members as their own felt values and that their interpersonal relationships are stabilized in reciprocal modes of behavior and mutual expectations. In natural groups, the pattern of statuses — that is, the relations between leader and follower — is not imposed from outside or through formal organization. Hence, as a general rule, leadership authority is not perceived as an autocratic imposition from outside. In our studies, we have records of such statements as "We have no leader," "We have no boss," and "We are all equal." Despite such expressions, members could be and were reliably ranked as first, second, third, and so on in terms of the frequency of their effective initiative in starting and in carrying on activities.

In brief, the natural group is a reference group for the individual member. It is a group whose acceptance the individual actively seeks, a group in which he aspires to amount to something, a group against whose yardsticks he gauges his own success or failure in relevant dimensions. The self picture of the individual is shaped, in no small part, by his reference group ties and by the yardsticks provided by his reference group.

Obviously, groups do not rise and function in a vacuum. The members' claims for attainment as well as the goals they are pursuing are relative to the setting of which they and their group are a part. That part of the setting that raises issues or problems for them in their moves toward goals is the sociocultural setting.

If an individual's attitudes and behavior are affected by his reference groups, and these groups, in turn, are related to the sociocultural setting of which they are parts, what does this imply for the study of human behavior? It means that, if we are seriously interested in the etiology of attitude and behavior, we have to study them amid influences ranging all the way from the society and the particular sociocultural setting of which he is a part down to the groups in which he actually moves and his own psychological make-up. Probably most psychologists would agree in principle that behavior is determined by such a gamut of influences. However, our

( 101) research traditions and specializations do not encourage us to include the gamut of influences in our research designs. Instead, we are likely to end up studying groups or personality variables, or surveying attitudes, or studying cultural influences in isolation-paying only lip service to the fact that they act, react, and interact, shaping behavior in an interrelated way.

In designing our research on natural reference groups, we took literally the need to include at least the essentials from the entire gamut of influences, and we therefore organized the research into three interrelated parts. First, low, middle, and high socioeconomic areas with differing ethnic composition were chosen for study so that we could specify differences and reach generalizations common to all levels. Second, we obtained data on acceptable and objectionable behaviors and goals among individuals, groups, and areas, both to compare an individual's attitudes with those in his group and with those prevailing in its sociocultural setting and to compare attitudes in various areas. Third, groups of four to twelve teen-age boys (13-19 years), selected on the basis of their frequent and recurrent association, were studied intensively for six to twelve months by means of such techniques as observation, behavior ratings by regular observers and independent observers, informal sociometric tests, and situational tests. The distinctive strategy of our research is that the subjects are not aware that they are being observed.

It would be perfectly understandable to ask, Why go to the trouble of dealing with such a variety of different kinds of data, especially when a lot of them are not psychological data? The answer is that all this is necessary if we are ever to achieve theoretical formulations and analyses that are valid. It is precisely because we in psychology have been so buried in fragmented segments of the many influences shaping behavior that our formulations are frequently shattered by real events and by attempts to apply our formulations.

Re-Formation of the Self System during Adolescence

The study of natural reference groups is particularly important in understanding the individual-group-society relationship during adolescence. It can also indicate the motivational basis for the intensified interactions with age-mates that are so typical of adolescents in highly differentiated, casually patterned modern societies. With a decade or so of life behind him, the growing child starts to develop physically into an adult. The

( 102) problems, events, and timing of adolescence vary enormously in different cultural and socioeconomic settings, but in nearly all societies, there are two sets of universals: (a) more rapid pace of growth involving profound biochemical as well as structural changes of the body, and (b) changes in the person's roles with accompanying shifts in responsibilities, activities, and ways of acting.

The changing body of the adolescent compels changes in his self concept. Self concept or ego refers to the constellation of attitudes that the person has formed relating his body and his individuality to other persons, groups, objects, values, and activities in institutions. Even if others were to take no note, the altering appearance of his body and its unaccustomed sensations tell the adolescent that he is changing. However, others do take note — his family, teachers, and friends expect new things of him; the mass media depict different ways of feeling and acting for his age group; he is expected to be more responsible and less dependent upon adults, at least in the conduct of routine activities and self care; he is expected to start to act, dress, stand, and walk in ways befitting the approach of adulthood. If the adolescent is to meet these expectations, he has to redefine relevant attitudes toward himself and others. In short, the fundamental psychological problem of adolescence is the general and important problem of forming and re-forming self or ego attitudes. Thus, adolescence is one of the best periods of development for studying an important problem in social psychology: attitude change.

In modern societies, however, the steps and procedures for attaining adulthood are by no means clear. For several years, the youth is neither child nor adult, neither dependent nor independent socially and economically. Adults are uncertain about the exact timing of steps and responsibilities he is to assume. Sometimes, when life has changed rapidly from when they were adolescents, adults offer guidelines so different from the paths open to the youth that the guides appear unacceptable. Furthermore, adult treatments are not consistent—the young person is treated as an adult one day, and subject to the restrictions and sheltered world of childhood the next. In short, to the extent that it is a prolonged transition with no clear and satisfactory procedures for gaining adult status, adolescence is a period of dilemma.

When faced with uncertainty and conflict about such central problems, any individual, of any age and at any place, searches actively for stable guideposts, some certainty, and some way out of the conflict. His choice

( 103) of guideposts is, of course, contingent upon his own selectivity and capability at the time and upon the alternatives available and salient.

Today, the social schemes for school, community, and leisure encourage the adolescent's choice of guideposts from age-mates. The adolescent discovers that there are others in the same boat, and his social interaction with other adolescents becomes more intense, more frequent, and more significant than in childhood. Adolescents in the United States are encouraged in moving toward age-mates by the emphasis on social activities, by pressure from parents and other adults who want them to be "in," by school programs, and by the varied mass media that deliberately feed on appeals to youth. The actual movement toward age-mates is also symptomatic of a general shift away from their strong ties of dependence upon adults — for the time being, at least, the standards that count, the opinions that decide whether I am normal, average, superior, or inferior are those within the domain of adolescents.

Problems of being accepted and of being recognized by age-mates become major concerns. To be neglected, rejected, scorned, or ignored at a time when one's self identity is changing is a painful experience to be avoided at all costs. When coupled with an earlier history of insecure ties with family and other groups, such experiences may impel an almost frantic search to belong someplace with someone, and at almost any price. What becomes important during adolescence, therefore, is closely tied with day-to-day relationships with age-mates.

Experimental Demonstration. The shift in reference groups from family to age-mates was studied experimentally in our research program by Prado (1958), who compared judgments by children and by adolescents on the competence of their own fathers and their best friends in a simple task. Prado selected twenty-five boys between eight and eleven years old and twenty-five between fourteen and seventeen years old. In order to minimize the possibility of getting boys who had unusually poor relationships with their fathers, he chose only boys who consistently selected their fathers as their most preferred parent. He singled out the boy's best agemate friend from sociometric data obtained at school.

Each boy came to the laboratory with his father and his best friend to perform the simple task of throwing a dart at a target. The task was arranged so that the exact score could not be determined by the subjects. Each boy judged his father's and then his friend's performance, using

( 104) scores from 0 to 24. The children and the adolescents were about equally accurate in their judgments, with an average error of 4.5 in judging their father's performance and an error of about 3.5 in judging their friend's performance.

The difference between errors in judging fathers and errors in judging best friends is an index of the comparative tendency to overestimate or underestimate performance by father and best friend. A positive difference would mean that the error in judging the father's performance was greater than that in judging the friend's and in the direction of overestimation; a negative difference would mean that the father's performance was underestimated, even when actual performance levels were taken into account. The findings were that children of eight to eleven years of age tended to overestimate their father's performance (mean difference, +2.35), whereas adolescents underestimated their father's performance compared to that of a friend (mean difference, -3.64). This experiment is one of several in our program demonstrating that systematic variations in a cognitive process (judgment, in this case) can provide precise indicators of the person's attitudes and affective ties.

Sociocultural and Ecological Characteristics of the Area

The consequences of the shift toward adolescent reference groups depend, in part, upon the facilities, the socioeconomic level, and cultural arrangements in which such reference groups function. These specific characteristics of the sociocultural setting are stimuli affecting the attitude and behavior of the subjects. We deliberately chose low, middle, and high socioeconomic areas of differing ethnic composition in order to reach generalizations common to all levels as well as to specify differences. Utmost care was taken to find measures to specify as exactly as possible the characteristics of the social areas in which the adolescents lived and moved. We relied especially on the Shevky-Bell Social Area Analysis (Bell, 1958; 1965; Shevky & Bell, 1955) to designate the ecological characteristics of each urban area. This analysis is based on the finding that the large number of separate items in census tract statistics tend to be intercorrelated and can, therefore, be reduced to three (or more) basic factors. Since certain items are better measures of these three factors than others, only a few of them are actually used in computing each of three indexes (Bell, 1965):

( 105)

Index of SES   Index of Urbanization or Family Status   Index of Ethnicity or Segregation
Rent   fertility ratio   race
Education   women not in the labor force   nativity
Occupation   single-family detached dwellings   Spanish surnames

Shevky and Bell standardized scores on each of these three indicators according to the range in Los Angeles in 1940. This standardization makes it possible to compare the rank of areas in different cities, even though the basic yardstick (Los Angeles, 1940) was arbitrarily chosen. For example, Figure 1 gives the distribution of census tracts in our research up to 1964 according to socioeconomic status and family or urbanization status. The location of an area in such a matrix can be used in plotting behavioral or attitudinal data, indicating both the social class (low, middle, high) and degrees thereof.

By exploring and mapping facilities of each social area and, in some cases, conducting surveys among residents, we specified still further the character of the area as it presents itself to the members of natural groups. Although these data are clearly at the sociological level of analysis, they do provide a framework for the study of behavior far superior to the "boxcar" designation of social class, as John Campbell (1967) has called the gross measures typical in sociopsychological research. It does make a difference whether a lower-class area is composed of poor families living in small shanties or working-class families crowded into apartments.

The ecological arrangements and sociocultural characteristics of the urban areas proved invaluable in understanding many activities and behaviors within adolescent groups. For example, the spacious homes and many recreational facilities in an upper-middle-class area were associated with activities and attitudes toward space that differed strikingly from those in a poor, overcrowded neighborhood, where privacy was difficult to obtain. The conflicts among groups of adolescents in culturally segregated, underprivileged areas frequently centered about the limited space and facilities, whereas intergroup conflict in middle- and upper-class areas involved prestige and power, concerns related to space and facilities in only tangential ways. Again, the total absence of interest in athletics in one group of adolescent boys would have been difficult to understand without the knowledge that they lived in a social area inhabited predominantly by first-generation Mexicans of peasant origin, for whom sports in the North


Figure 1.

Figure 1. Urban areas from two cities studied in the program plotted according to socioeconomic rank and urbanization (termed family status by Bell). Each dot represents a census tract in which members of a given group lived. Each group was studied intensively. Axes indicate index numbers in Shevky-Bell Social Area Analysis. The index numbers are based on average percentages of several census measures (see text) which are standardized by weighting coefficients that adjust for the range of percentages obtained in extensive analysis of census tract statistics by Shevky and Bell, then converted to index numbers. Thus, index numbers can be compared between cities as well as within cities. Horizontal axis indicates socioeconomic rank. Vertical axis indicates increasing urbanization of family life-style (i.e., the higher on the axis, the higher the combined average for percentages of multiple-family, attached dwellings, of women working, and of low birth rate). Black dots indicate segregated areas with populations of ethnic status low on the social distance scale in the United States (Mexican Americans, in this case). Three areas of highest socioeconomic rank (upper right) housed segregated white populations, and intermediate areas indicated by white circles were mixed (not indicated in Shevky-Bell notation).

American sense were not normatively defined as activities associated with masculine identity.

Self Radius and Goals Prevailing within the Settings

The problem of integrating sociocultural data and behavioral data based on individual responses has plagued social psychology. For sociopsychological analysis, specification of major characteristics of the setting is necessary, but not sufficient. Certain aspects of the sociocultural setting have greater salience than others for individuals of a particular age level studied in their natural groups; literature on the subject shows that adolescents are particularly tuned to others of their own age. Accordingly, special emphasis is given in our research program to data on the bounds of what are acceptable and objectionable behaviors within the group, and on the goals prevailing in each sociocultural setting among representative

( 107) samples of teen-age youth. Data from different areas can be compared to see what is common to them all and in what ways they differ. Data from Social Area Analysis provided the baseline for assessing values and goals prevailing among adolescents in the areas, and the data thus obtained, in turn, served as the baseline in assessing the attitude and behavior of specific individuals participating in small groups as members.

The Shevky-Bell indicators provided the basis for determining the social rank of areas where the self radius and goals prevailing among adolescents were assessed. Representative samples of high school students in the areas provided the data on the values and goals prevailing within each area, obtained through a questionnaire administered as a public opinion survey. The responses were treated as indicators of what youth in each area regarded as socially desirable reactions in the school situation. They permitted the estimation of what was valued, the level of goals set by adolescents in each area for the near and more distant future, and what were the minimum and maximum limits to the youths' own conception of success. The designation self radius refers to such limits as defined through various questionnaire items.

Impact of the Larger Society. Despite the rather sharply graded socioeconomic and ethnic stratification in the United States, there were marked similarities among the youth in different urban social areas. Though differing in appearance, specific styles of dress, and sometimes language, they all bore the unmistakable stamp of the larger culture, with its stress on individual success in private and material respects. The ingredients of individual success desired by nearly one hundred per cent of every sample of high school students included automobiles, comfortable homes, spending money, and at least a couple of hours a day "to do what I want to do."

Table 1 shows that responses of students in social areas of low, middle, and high socioeconomic rank in one city who responded to the question, "If I had my way, I'd like to live in a neighborhood in the X part of the city." It shows the percentage responding with the same location where they currently live, the "north side" of the city (which was undeniably the most comfortable), or some other location. As the table shows, the desire to stay in the same location decreased regularly from the high- to the low-ranking neighborhood, and the desire to move north increased (the high rank sample already lived there). If all had their way, almost all of them would be living in the north side of that particular city. The impact of the larger society was also apparent in other respects — notably, the great im-


Table 1. Area of Residence Desired by Male Adolescents Living in Social Areas of Differing Shevky-Bell Socioeconomic Rank and Ethnic Composition
Area of City Desired High SES Segregated, Anglo Middle SES, Mixed Low SES, Segregated “Latin”
“Anglo” “Latin”

Same as present















Total N





* North was the area of residence of the high rank sample (chi-square test between “same” and “all other” responses was significant (p < .001, 3 df).)

- portance assigned to at least a high school education and aspirations toward occupations of higher status.

Limits of Success Determined by the Setting. Since high school students are likely to give responses on a questionnaire that they consider socially desirable, it is interesting to see that what is socially desirable does differ according to socioeconomic rank of the area. As other research would lead us to expect, the absolute level of goals for future occupation, education, and income differed systematically from one area to the next. Our data yielded the additional finding that what was conceived of as "success" in these respects also differed. For example, Table 2 gives the average

Table 2. Reference Scales for Subsistence — Comfort and Goals for Weekly Income (1959) of Adolescent Males Living in Social Areas of Differing Shevky-Bell Socioeconomic Rank and Ethnic Composition
Response * High SES Segregated, Anglo Middle SES, Mixed Low SES, Segregated “Latin”
“Anglo” “Latin”
“Barely enough to live on”
Median $59 $55 $41 $35
C.L. 52—77 51—59 34—55 31—46
“To be really well off”
Median 318 172 94 83
C.L. 239 — 455 142—207 78—150 70—100
“I want to make in 10 years”
Median 229 142 98 85
C.L. 184—300 122—175 75—131 77—97
* Medians were computed because distributions of response were markedly skewed. CL refers to confidence limits of the median, which vary with both dispersion and number of cases. Despite overlap of confidence limits for response “barely enough to live on,” a median test of the response distributions was significant (p < .001, 3 df).

(109) median) responses to a series of items concerning how much weekly income was needed to "just get along," was needed to be comfortable, and was desirable for him within ten years. Spanish-speaking samples were living in areas of low and middle socioeconomic rank. As the table indicates, the radius within which one sets his future aspirations differs from the high- to the low-ranking areas and is also affected by membership in a cultural minority deprived of opportunity. Psychologically, the adolescent's reference set established limits for the radius of the self in these respects.

Nevertheless, relative to their fathers' attainments and to their own conceptions of success, aspirations for upward mobility were common in middle and lower rank areas. Table 3 shows that approximately 40 per cent of the youth in low- and middle-ranking areas of one city aspired to occupational levels higher than their fathers'. This upward shift resulted from less desire for manual labor and greater desire to reach the upper occupational categories.

Table 3. Discrepancy between Adolescent's Desired Occupation and Father's Occupation
Occupational Rank Rank of Area
High Middle Low
1. Highest professional, top managerial - 0.8%  + 2.5% 0.03
2. Professional, business, technical -12.5 20.3 12.1
3. Lower professional, small business 11.6 10.6 17
4. Sales and Clerical 10.1 8.6 9.7
5. Skilled -7.6 -25.6 -4.9
6. Semiskilled -0.8 -10.3 -22.6
7. Unskilled 0 -6.1 -14.3
Percentage differing from rank of fathers' occupation 21.7% 42.0% 41.8%

Similarly, in response to inquiry about the amount of education "necessary for a person who wants to do the things I want to do" and the amount the subject himself desired, the aspirations in low, middle, and high areas differed systematically. The median was completing high school in low-ranking areas, some additional training in middle-ranking areas, and college in the high-ranking areas.

Greater Heterogeneity of Goals in Middle and Lower Rank Areas. The data on educational aspirations serve as a point of departure for another

(110) important finding: The values and goals that prevail among adolescents in lower- and middle-class neighborhoods are more heterogeneous than those in upper rank areas. Figure 2 consists of data on educational goals presented as a frequency distribution of the responses (number of years desired) for low, middle, and upper rank samples. The responses of youth in the upper rank area were most homogeneous, about 85 per cent desiring four or more years of education beyond high school. Although the typical response in the lower-class area is for a high school education, a sizable minority aimed at college. Certainly, there is little hint here of the supposed derogation of school that once was believed typical of the lower class. Similarly, the distribution in the middle rank area reveals considerable heterogeneity, even a bimodality.

In this and several other respects, the findings showed that youth in upper rank areas were most like-minded. Coupled with the fact that youth in such areas are also more frequently and extensively involved in formally

Figure 2.
Figure 2. Percentages of adolescents in urban areas of different socioeconomic rank expressing educational goals for school attendance from eight or fewer years to more than eighteen years

(111) organized activities in school and community, this finding suggests that we might also expect to find less variety in the impact of age-mate reference groups in high rank areas. The intensive study of natural groups in lower and middle rank areas supported the conclusion that the impact of the group varied tremendously in terms of whether or not the group norms supported or were contrary to the values prevailing in and promoted by the high schools, law enforcement agencies, and other community agencies.

Intensive Study of Natural Adolescent Groups

The focal aspect of our program is the intensive study of natural groups. Since the program is sociopsychological, its procedures and data are concentrated on selected small groups and their individual members. Hence, the first requirement of the research design was the selection of clusters of teen-age boys (13-19 years) on the basis of their frequent and recurrent association in specified settings. Each group of four to twelve is studied intensively for periods of six to twelve months through a combination of techniques, including observation, behavior ratings by regular observers and independent observers, informal sociometric tests, and situational tests. The subjects are not aware that they are being observed. People do act differently when they know they are being watched and sized up; this is particularly true of adolescents, who are on guard in their relations with adults. Behaviors in the groups studied have included both socially acceptable and socially objectionable activities. However, groups were selected by their regular and recurrent association, and not by whether the members behaved properly or misbehaved. Thus far, we have completed the intensive study of four dozen groups in eight cities through the study cycle described below; we have partial data on a much larger number of groups.

The study is systematic in several respects. First, it is focused on obtaining data pertinent to hypotheses about the individual-group relationship. Second, it collects data on individual behavior or group members and on group processes in a systematic manner, using defined procedures in a definite sequence. Third, it seeks to relate these data systematically to the sociocultural and ecological findings, using the latter as the baseline to plot the attitudes and goals of group members and thereby viewing their conformity-deviation relative to others in their own area. Our major hypotheses and a string of auxiliary hypotheses in the present research were presented earlier (Sherif & Sherif, 1964). Here, we shall concentrate on two

( 112) major areas of theoretical interest: (a) To the extent that a set of norms is stabilized in regulating activities frequently engaged in, these norms become binding for the individual member; however, his conformity to them depends upon his status in the group. One indicator of stabilization of a set of norms is the extent to which members abide by them without threat of sanctions. (b) To the extent that a pattern of reciprocities is established as role and status differentiation among members, the behavior of the group members becomes more predictable. In other words, the more tightly knit the group, the more the interpersonal relations of group members and their attitudes toward outsiders become governed by it.


Our cardinal research strategy in studying groups is to accommodate the properties of every group as it actually forms and functions throughout the six- to twelve-month study cycle. We are acutely confronted with this issue from the very beginning as we start pinpointing a group for intensive study. First, an observer — a young adult college student or graduate student —is chosen to study one group. His first job is to master the instructions about steps and methods of observation (Sherif & Sherif, 1964, Appendix). In general manner, physical appearance, and speech, he has to fit into the area he is to study. Thus, a Spanish-speaking observer goes to a Mexican neighborhood, a Negro observer to a Negro neighborhood, an observer of upper-class background to a high rank neighborhood.

Next, he singles out a group — at first, exclusively on the basis of observed frequency and regularity of their association in the area (on street corners, in recreation areas, vacant lots, soda fountains, pool halls, skating rinks, and other hangouts). When he observes a minimum of four or five boys associating repeatedly, this cluster becomes a possible group for study.

The observer's first attempts are directed toward establishing his presence in the area as a harmless or neutral outsider in the eyes of the individuals. He has to develop a clear reason or pretext for being there, because there is a universal human tendency to categorize or place another person whom one meets for the first time. In other words, the observer himself is being observed, and in order to foster a categorization conducive to studying a natural group, he must be placed by the members as a harmless person who is not an authority figure or researcher or a reformer.

When, and if, he succeeds in establishing contact with the potential

( 113) group, he begins intensive study. This consists of regular reports on each observation period, status rankings made on the basis of effective initiative and patterns of deference exhibited by each member, and ratings by an independent observer to cross-check the reliability of observer's ratings over a time span. Later in the study cycle, the observer's outside view of the status pattern is further checked against sociometric choices secured informally and privately from each individual to reflect perceptions from inside the group. The cycle also includes the observation of products peculiar to the group — recurrence of preferred activities, labels and nicknames, common practices and preferences, and reactions to cases of deviation.

The strength of group ties or solidarity (cohesiveness) of the group is checked through behavior of members in test situations and indicators of the effort they are willing to exert to associate with fellow group members. For example, such indicators are the alternatives chosen by the member when he is caught between staying with his group in a planned activity or getting away from it for a date, a family obligation, church, or school. Another indicator is the geographical spread of the members' dwellings, which reflects the effort and even sacrifice each has to make to associate with his fellow members rather than with others physically closer to his home. A third indicator is the frequency with which members know each others' whereabouts when they are not together.

Only after the observer has established close rapport with members, which usually takes several months, does he interview and directly question members to assess the natural history of the group's formation and to secure case histories of individual members in the neighborhood, including information about their homes, families, school records, and so forth. After a group is pinpointed, it usually takes from six months to a year to complete the cycle of study for it. The entire cycle adds up on the average to about a hundred observation periods for a single group.

The various data-gathering methods and procedures are introduced at a time appropriate to the degree of rapport developed between observer and the group members. We do not force upon members any direct questioning, interviewing, or other procedures which would be considered an imposition or an unwarranted intrusion into their privacy. Also, a special effort is made to keep members from becoming aware that the observer—who is hanging around as a harmless, well-wishing, older friend —is studying them and rating their behavior. Even with all of these precautions and a hands-off policy with regard to the planning and execution of group

(114) activities, the observer must maintain an uncritical attitude regardless of what they do. In spite of all such precautions, the establishment of rapport did not prove possible in every case. The reasons for such failures or for partial success are perhaps equally as important as the reasons for taking the precautions.

Groups are not formed for the benefit of outsiders to study them, including psychologists. They are not formed for a detached worker to reform them. They are not formed to tolerate criticisms, advice, and especially unfavorable evaluations by outsiders. They are formed in the process of interaction over a time span for satisfaction of members' common motivational urges, regardless of what these urges may be. Therefore, groups have designs, plans, and ends of their own, proportional to the extent that they are at odds with the established routines and channels, with other groups and institutions in their surroundings. They have secret designs and plans, and resist intrusions into their privacy, including intrusions by researchers and reformers. Invariably, in groups we have studied, the most difficult hurdle has been the initial period of secrecy and resistance to the observer, which he may not always overcome despite several months of persistence. When he did not, the study of such groups could not be carried through the entire cycle. We suspect that, even though the study cycle could not be completed, the walls of resistance and secrecy indicate tightly knit groups with highly stabilized organizations and binding commitments shared within the bounds of their particular membership.


In every case, for a period of several months the observer himself is under scrutiny by group members before they tolerate his presence in locations and spheres of activity which are the group's exclusive business. The critical phase toward establishing rapport is conceptualized as "the observer being observed." The members ask the observer point-blank who he is, what his connections are, and why he is hanging around; over the course of time, they check up on these points before they start opening themselves up — a luxury denied the observer himself at this initial phase of study. This period puts challenging and vexing demands on the patience, ingenuity, and skill of the researcher.

Gauging the degree of rapport attained at a given time should not be merely a matter of intuition. We have devised a combination of indicators of rapport for the timing of the several procedures in the study cycle.

( 115) These indicators include the observer's success in finding out the places where the group congregates when they are not in plain public view and their toleration of him in these places, the degree of intimacy of the activities they talk about freely or discuss in his presence, and the extent to which members permit or even welcome him into activities which they consider strictly private.

Why is degree of rapport important in decisions about research procedures that are appropriate at a particular time? It is not just a matter of whether or not one can collect data, for this is usually possible, but a matter of whether the data collected by a given procedure at a given time are valid as indicators of group properties and members' private attitudes. Strong emphasis should be placed upon the basis for developing rapport in order to secure valid and comprehensive data on group processes. This basis includes the assurance developed over time that the observer will not reveal private or secret activities to authorities and, in fact, has no connection with them. Such assurance is impossible for detached workers or others with official commitments to societal agencies, but it is necessary to ensure valid data about interaction among adolescents that is not deflected by an adult.

Patterns of Interpersonal Relationships within Groups

In this section, we shall concentrate on those indicators that are most useful in making accurate predictions about members' behavior. They can be accurately predicted because they are derived from the interaction process among individuals who really count in each other's eyes while they engage in activities that are important to them. These indicators include planning, discussion, and execution of actions that the members consider private or secret, and would not have exposed to any adult with authority. The first indicator is the member's position in the pattern of interpersonal relations of the group. The second pertains to the norms of the group. These two aspects of group process are closely related.

Our general finding is that associations of the sort we have studied are patterned affairs. Being a part of regular and recurrent interactions among age-mates who congregate of their own choice means regulating one's behavior with that of other group members. Over a period of time, the interactions among individuals become patterned and more predictable, so that they can be reliably studied in a variety of specific situations. Interactions follow various patterns, but the most predictive of behavior

( 116) is the power dimension, as revealed by observers' ratings of effective initiative during interaction.

In our earlier studies of group formation and intergroup relations in summer camp settings, we had observed certain regularities in the patterning of interpersonal relations and power within a group (Sherif & Sherif, 1953; Sherif et al., 1961). It is significant that a similar pattern emerges in the perceptions of the members of a natural group, as indicated by their sociometric choices, and in the status ratings made by an observer over many observations when all members are present.

The member's position (rank) in a developing power structure is his status in the group, a differentiated aspect of his role that is defined in terms of the relative effectiveness of his actions in initiating, making, or approving decisions; coordinating interactions; and invoking sanctions for deviant actions by others. His status is further defined in terms of the evaluations of other members. The member's status, therefore, reflects a power dimension of role relations that is not identical with his ability to influence another's behavior in any situation at all. Nor is status identical with prominence, expertise, or the degree to which the person is liked. Power is manifested in situations where influence can be implemented with sanctions.

The referents of status and its significance in predicting members' behavior can be clarified by a diagram representing the process of group formation over time, as defined by the power or status dimension (Fig. 3). The status dimension in the diagram refers to the relative effectiveness of members in initiating activities over time in a variety of specific situations. At the top (Time a) are two collections of individuals with no previous history of interaction, but homogeneous enough socioculturally to hold other bases of differentiation constant. The circles indicate that, at this time, the individuals' relationships cannot be reliably ranked according to effective initiation of activities or of controlling interaction from one situation to the next; instead, the ratings of each person's effective initiative are different in various activities.

The course of status stabilization diagramed in the figure is based on the cumulative findings for six groups in a summer camp, formed experimentally by placing unacquainted, similar boys eleven to twelve years old in problem situations with highly appealing goals that could be attained only through coordination of action. The diagram represents two groups in the process of stabilization, to indicate that the particular pattern of organiza-

( 117)

Figure 3.

Figure 3. Diagram of the formation and gradations of status structure over time from initial contact (top) to a stabilized group (bottom) in two illustrative cases. Stabilized positions (triangles) are based on agreement among several methods for ranking members according to their relative effectiveness in initiating plans and actions during prolonged interaction in naturalistic settings when facing problem situations of high motivational appeal. Note that Case A develops more vertical status levels than Case B, which has fewer differentiated status levels. Such differences were observed in the experiments that provided the data for constructing this diagram.

-tion, the steepness or flatness of the hierarchy, and which individuals occupy what positions do not follow any set, predetermined form.

In the second frame at Time i the lowest and the highest positions are represented by triangles. The triangle stands for positions that can be reliably rated from one activity to the next and from one day to the next. The top position typically stabilizes earlier than other positions, followed by the lowest status positions.

At Time j in the figure, observers are able to agree on the positions most members occupy from one situation and from one day to the next, with the exception of the middle of the organization. Again, this is a typical finding for groups studied in both experimental and field conditions. In part, the continued flux in the middle ranks may reflect attempts by those in the middle to improve their standing or to align themselves with those of higher status. At Time n in the figure, the status relationships are stabilized; this is revealed both by all observers' agreeing on the status structure

( 118) and by members' perceptions of their relative standing, obtained through sociometric choices.

The speed or rate of stabilization diagramed in the figure will vary. In the particular camp experiments on which the diagrams were based, the groups stabilized within about a week while members lived together continuously and regularly engaged in highly appealing activities requiring coordinated action to attain goals within their own groups. Other investigators have reported discernible beginnings of group structure among individuals meeting in the same location for similar activities within three to five meetings lasting a few hours. Environmental events affect the rate of stabilization at least as much as internal relations of members. The stability achieved is sensitive to the introduction of new members, to changes in location and facilities, and to outside threat or emergency. Among the most solidary groups in our field study were those whose members had been forbidden by adults to associate, either as individuals or as a group. Thus, both a lower-class group of troublemakers and a high school fraternity ordered to disband were continuing their groups underground.


The general finding, then, is that regular and recurrent interactions among adolescents with common goal directions invariably become patterned through differentiation of member roles and status positions. Accordingly, the interactions among individuals become more predictable from one situation, one activity, and one time to the next, indicating recurrent regularities in attitude and behavior. The interactions are patterned in various ways, the dimension most predictive of behavior in important activities being the power dimension. Who gains what position remains an important problem for study.

One invariant in the stabilization of organization concerns the perception and judgment of the organization both by the members and by the observers. Theoretically, the important generalization about the stabilization process is that perception of the organization is polarized by the extreme positions at the top and bottom. Thus, in social organization, we are dealing with a general psychological phenomenon, namely, end anchoring. The end anchoring phenomenon refers to the finding in judgment experiments that extreme stimuli in a series (weights, frequencies, lengths, or position) are identified, learned, and judged more quickly and more accurately than intermediate stimuli. In serial learning and perceptual recog-

( 119) -nition in serial presentations, end anchoring occurs as a function of order of presentation, with the earliest and the last stimuli in the series serving to define the set (Harcum, 1967). De Soto and Bosley (1962) reported end anchoring in paired associate learning, in which subjects were required to learn names of unknown students paired with their college class (freshman, sophomore, junior, senior). Learning was faster both in number of errors and number of trials to perfect recall for the lowest (freshman) and highest (senior) classifications.

As Figure 3 shows, end anchoring was observed in the stabilization of six experimental groups. In the observation of several dozen existing groups of adolescents in natural field conditions, the observers' ratings of status over periods of time ranging from six months to a year were analyzed, with the invariant result that the ratings were more consistent for the highest and lowest status positions, with greater variability of rating in the intermediate ranks.

Figure 4 presents the composite findings on the ratings of a number of observers, each rating a different group. (Separate analysis for an individual observer of a particular group over time is presented in Sherif & Sherif, 1964.) The separate curves in Figure 4 represent averages in the vari-

 Figure 4.

Figure 4. End anchoring in observers’ judgments of status in natural groups of adolescents over time. Mean change in ranks assigned by observers according to observed effective initiative (status) during blocks of ten observation periods early in study cycle (Time a) to late in th study cycle (Time n).

(120) ability of ratings for separate blocks of observation periods (ten periods in each block) from the observers' initial encounters with the groups (Time a) to Time n, after they had observed for some months.

As Figure 4 shows, the most extreme representatives were recognized most readily and used as standards in judging others. It is important to note one difference from the end anchoring that occurs more typically in judgment or learning of a neutral series of stimuli. As a social structure, a group is affected by the superior anchor (high status) to a greater degree than a physical dimension would be. That is, the leader position exerts a greater anchoring effect than the bottom positions, both on the judgments of an observer and on promoting the stabilization of the structure (note that variability in rating lower positions is greater than that in rating higher positions). In general, the observer's variability in ratings decreases over time (Time a—Time n). This trend occurs, of course, when no drastic changes occur in the status structure during the time period.

As an important correlate of the end-anchoring effect, observers are more confident of their ratings at the extreme positions (see Fig. 5). The observers regularly rated their own confidence in the ratings they made on

 Figure 5.

Figure 5. Mean confidence of observers associated with end anchoring in their judgements of status. Observers’ self ratings of their own confidence from altogether confident (6) to not at all confident (0) when ranking members according to status in natural groups early (Time a) and later (Time n) in the study cycle.

(121) effective initiative immediately following each observation period on a 7point scale ranging from altogether confident to not at all confident. As Figure 5 shows, observers' confidence in their ratings increased over time, but was invariably highest at the extreme status positions, especially at the high end.

The finding that end anchoring is a general phenomenon of human judgment, both in laboratory experiments with neutral stimuli and in observations of group formation and group functioning, is all the more significant because it is not confined to trained observers. The same phenomena are detected in the participants' own perceptions of the group structure, as obtained through such sociometric questions as, Who gets things started? and Who gets things done? Consensus on choices is greatest for those in highest status positions (as independently assessed by independent observers) and those at the lowest positions (who receive few or no member choices on such items). In short, the stabilization of a hierarchical role and status structure is an invariant result of interaction over time among individuals facing common problems. Adolescent interactions are no exception: Choosing groups for study entirely on the basis of the recurrent, regular interactions among individuals, we have found that interpersonal relations among them are patterned and more or less orderly affairs, no matter how casual they may appear to the outsider.

The important result, from the psychological point of view, is that the status structure and the norms of the group provide a remarkably accurate basis for predicting the behaviors of individual members. This generalization has been tested using laboratory-like methods. The techniques are based on the general principle that a person's attitudes, expectations, and motives are revealed in his cognitive processes (e.g., in his judgment) when he faces an unstructured situation relevant to them. In other words, perception or judgment can serve as an indirect behavioral indicator of the person's attitudes, provided that the stimulus situation lacks objective structure in a relevant dimension. This principle had previously been demonstrated in experiments on judgments of future performance as affected by interpersonal attitudes (parents and children, husbands and wives, C. Sherif, 1947), by the mutual expectations of friends and rivals (Harvey & Sherif, 1951), by role and status expectations within small groups (Sherif, White, & Harvey, 1954), and by hostile attitudes toward a rival group (Sherif et al., 1961).

( 122)


Earlier experiments suggested that members' appraisals of one another can serve as an unobtrusive or indirect method for studying status and role relations in a group. This possibility was explored further in a study by Koslin, Haarlow, Karlins, and Pargament (1968), which secured judgments on four different tasks from members of four groups. Specifically, these investigators asked whether differential expectations of performance by individuals with differing status in a group would be generalized across tasks and activities. They predicted that consistency in overor underestimation of performance would be greatest for individuals occupying the highest and lowest status positions, with greater variability from one task to the next for judgments of members with intermediate status. This hypothesis was derived from the general principle of end anchoring.

In an athletically oriented camp for boys, four groups were observed over time, using the unobtrusive procedures described earlier. Each group consisted of six to eight boys between eleven and thirteen years of age; each group lived in a separate cabin (N = 29). The observer rated the effective initiative displayed by each boy in a variety of situations.

The four tasks used to tap expectations of performance were as follows: (a) A sociometric questionnaire administered to each boy individually on the pretext of improving the camp. (b) The height estimation test. Stick figures were presented to each boy with the instruction to imagine each figure as a member of his group, then to indicate where the top of his own head would reach on the figure. (c) The rifle task. Each boy shot one bullet at each of four targets placed fifty feet away. When a bullet hit the target, it disappeared. Other members observed his performance and recorded their judgments of his performance. (d) The canoe task. Each boy paddled from a bridge around a buoy and back to the bridge, whereupon other members recorded their estimates of the time required for performance. In the last two tasks (rifle and canoe), group members were already familiar with one another's proficiency before the estimates were obtained. The differences between estimated and actual performance were adjusted, therefore, to take account of level of skill.

Significant correlations were obtained between judgment variations in each pair of these four tasks and judgment variations in each task correlated with observers' ratings of status. The highest association between

( 123) observed status and judgments was obtained when a multiple correlation was computed, using observed status as the criterion variable. The degree of relationship between performance expectations on the four tasks and observed status is indicated by the multiple r of .79 (p<.001).

Separate analyses of results for each group were made to evaluate the second hypothesis —namely, that end anchoring of judgments would be revealed through greater consistency (less variability) across the four tasks for members with highest and lowest status positions. Since the measurement units for the tasks were not comparable, the scores on each task were ranked within each group from highest to lowest. Then, the variation among each member's ranks on the four tasks was computed. This variance among ranks was correlated with status rank based on the observer's ratings using the statistic E (curvilinear association). Only the smallest group (Group II) in Figure 6 did not show a significant departure from linearity, and, even for that group, the curvilinear relation accounted for 65.6 per cent of the variance in the ranks.

 Figure 6

Figure 6. End anchoring in judgments made by members of natural groups I—IV of the performance by fellow members in four tasks (combined). Mean variance in ranks of members with differing observed status (high to low effective initiative on base line) when ranked separately on each of four tasks according to mean errors in judgments made by other members of their performance. Reproduced from Koslin et al., by permission of the American Sociological Association and the authors.

(124) Koslin et al. concluded that their findings support the predicted relationships between sociological status in the group and member expectations on several independent tasks, suggesting a generalization of expectations for the member across tasks. However, this generalization is much greater — and hence expectations are more consistent — for high and low status ranks, as expected if end anchoring occurs in perception of group structure.


Is effective initiative simply a function of the individual's attempts to respond adequately to the demands of the immediate situation, or do members' expectations on one another based on past interactions affect the exercise of power in a group? The answer to this question has important practical implications: for example, if one wants to influence an adolescent group by passing along information, which member will be most effective in influencing group members, one with high status or one with low status? An experimental study by MacNeil (1967) indicates that the answer to our question depends upon the stability of the group structure. Six natural groups of adolescent boys were selected on the basis of observations over periods from two to seven months. Using the reliability with which observers ranked each individual's status (top to bottom) as the primary criterion, several judges rated the stability of the groups, three being rated as high and three as low.

In each group, a member of high status and a member of low status were selected to be indoctrinated in an experiment with an arbitrary norm for judging highly unstructured situations (the autokinetic situation and the number of holes in a target made by a shotgun blast). In either case, the member was offered the chance to earn five dollars by substituting for a college student who had failed to appear for an experiment on the "human mind as a calculator." He served with a planted subject who distributed his judgments around an arbitrary mode and within a range much higher than was typical in the judgment situation. Since the judgment situations were highly unstructured, there was little difficulty in indoctrinating the group member with this arbitrary norm.

The question asked in the experiment was, What would happen when the indoctrinated member (either high or low in status) subsequently judged the same situation with other members of his group? Would the other members accept his judgments, knowing that he had previous expe-

( 125) -rience in the experiment? MacNeil predicted that the answer would depend upon whether or not the group was a highly stable one and upon the status of the member.

Figure 7 shows that the effects of the indoctrinated members on judgments of other members of their groups varied markedly. The high status members of well-stabilized groups were very effective in influencing other members and maintaining the indoctrinated norm over five sessions, whereas low status members in the stable groups were scarcely effective at all —almost as if the other members had discredited their judgments before they started and evolved their own norms in the situation. In the less stable groups, the indoctrinated norm had a significant effect, regardless of whether the member was high or low in status. Here, the experimental situation itself became more important in determining the outcome: the fact that one of the members had previously served in the situation gave some credibility to his judgments, as indicated in the higher medians evolved by the less stable groups. On the other hand, in neither case was

 Figure 7.

Figure 7. Median judgments in two tasks (combined) made by members of natural adolescent groups with high (H) or low (L) stability plotted for five sessions over time according to the status (hi or lo) of member who received highly arbitrary information in preceding session. Arbitrary information consisted of judgments of extraordinarily great magnitude nade by a supposedly experienced subject. Ordinate gives medians for combined judgment scores on two tasks, adjusted for scale unit differences. The higher the score the more effective was the member in inculcating the arbitrary judgments that he had acquired in the earlier session (not shown) with the “experienced subject.” Reproduced by permission of MacNeil, 1967).

(126) the member of a less stable group so effective in establishing and perpetuating the arbitrary norm as was the high status member of a stable group.

Conformity and Deviation of Group Members

The members of natural groups are "significant others" to one another. These reference groups constitute the significant and proper locus for studying patterns of conformity and deviation of their members. When we can locate an individual as a member of a reference group with specified properties, we have a basis for predicting what values he will conform to and which he will deviate from with little or no sense of guilt and with few or no pangs of conscience. More than that, if we know his position in the pattern of interpersonal power relations in that group, we have an effective basis for predicting the extent to which the values or norms of that group are binding for him personally, without threat of sanctions or other direct pressures from others.

Because much controversy has raged over the concept of "subcultural" variations in different neighborhoods and in different groups engaging in delinquent activities, it is important to note some features of the normative aspect of any group, whether delinquent or not. First, normative regulation is not to be found in all spheres of behavior or all activities. The convergence on customs and standardized outlooks by group members, with implied or actual sanctions for deviation, applies especially in matters of some consequence to the group and its members. Second, by their very nature, the norms of a group permit, even recognize, the play of individual differences. The norm defines a range or latitude for acceptable behavior, as well as a latitude of objectionable behavior to be met with disapproval or other sanctions. These two considerations become immediately obvious in actual observation of adolescent groups, whether delinquent or not.


We have found that membership in natural reference groups is not binding to the same degree for all members, or in all matters. The latitude of acceptable behavior defined by a group norm, and the extent to which the norm itself is binding, have been determined through test situations contrived in the natural setting. The range, or latitude of acceptance, within which norms are binding depends both on the behavior in question and

( 127) upon the status of the individual in question. This generalization can be summarized briefly as follows.

The latitude of acceptable behavior, defined by group norms, varies according to the importance of the activity for the group. In matters affecting the identity, maintenance, and perpetuation of the group as a unit (such as loyalty in the face of outside threat or intrusion), the latitude for binding adherence to the group is restricted, and the bounds are sharply defined. Deviation in such matters calls forth strong and consensual sanctions from others, even separation from the group.

But even in matters of central and common concern, the norms are not equally binding for all members. The higher the individual's position, the narrower the latitude of acceptable behavior for him in such matters. The leader is expected to be exemplary in these respects, particularly in relations with outsiders. On the other hand, the leader is not placed under so great pressure to conform in matters of lesser importance, especially those occurring strictly within the confines of the group. For example, he can commit fouls in sports not to be tolerated from others, and he can offer insults not to be endured from those of lower status. On matters of minor importance, particularly those concerning the members when they are within the group, the leader's behavior varies within much broader bounds.


Of considerable import here is the question of how binding the group is for the individual adolescent when he is not in the presence of other members. Our observational data showed that once the individual has adopted the standards of his reference group, he is remarkably impervious to attempts by teachers, parents, or other age-mates to sway him from them. For example, one boy whose hair was considered a disgrace by his father and teachers and a joke by other teenagers refused to have it cut until members of his own group convinced him that his hair was a little beyond the bounds even of their norm.

This general question was examined experimentally by Pollis (1964; 1967), who selected members of natural groups and age-mates who were not friends to serve as experimental subjects. In the first session of the experiment, group members judged an unstructured stimulus (auditory) until they converged upon a common range and mode of judgment. Other subjects had the same experience either alone or in the company of an

( 128) age-mate with whom they had no previous relationship. In a second session, the group members were separated and placed with two other subjects who were comparative strangers. In this session, each of the three subjects already had a norm or reference scale for judging the situation, but each had a very different reference scale. Further, each subject differed with respect to the conditions in which he had formed his standard —either by himself, together with a comparative stranger, or with fellow group members.

Pollis asked the simple question, Who would be most influenced by the divergent judgments of others in the second session and who would, in turn, exert the most effect on others? The answer was that the individual who had evolved his own standard, individually, was most influenced by others and was least effective in swaying others; the group member who had formed his standard with fellow members adhered most closely to his norm and was more effective in influencing others. Those who had formed standards with comparative strangers were intermediate in both respects.

This simple experiment confirms a finding that appeared over and over in our observations of natural groups in the field: Proportional to the stability of a group, its norms are binding to members even when they are not face-to-face. As Pollis's experiment suggests so clearly, this adherence is not blind conformity but a product of interaction with high affect for people who really count for one another. Unless this emotional nexus is clearly understood, there is little chance for adequate analysis of the phenomena of conformity-deviation during adolescence.

Concluding Remarks

The properties of naturally formed reference groups place limits upon the procedures that the researcher can use and upon the sequence of their application, if he wants to collect valid data on adolescent attitude and behavior. Much more effort and patience are required than in laboratory research. But within the approach outlined here, measurement that is both precise and reliable comes within the grasp of field research. The precise methods of laboratory assessment can frequently be adapted for use in the much more complex arena of natural settings, and the laboratory, in turn, can be used effectively in conjunction with field research when the major essential variables are incorporated into the experimental design.

The enormously complex problems of the individual-group relation-

( 129) -ship during adolescence require the marriage rather than the divorce of theory and research methods from psychology and sociology. When small groups and their members are studied relative to their sociocultural settings, and the setting is studied relative to the group as a part of the setting, the dichotomy between the study of interpersonal interaction and the study of culture and social organization, which is still so prevalent, will disappear. What are considered the "psychological" and the "sociological" studies of social behavior will supplement each other, instead of being monopolistic preferences of their respective disciplines. On the psychological side, the contributions of particular individuals, for good or for evil, can be studied to any desired degree of elaboration within their appropriate stimulus settings. These settings are patterned affairs: They are the individual's reference groups and the sociocultural surround of which these reference groups are parts. The contribution of his personal skills and qualities to the behavioral outcome is not abstract —his unique individuality is seen in the ways he interacts with others in activities and situations that are important to the individuals in their concerted undertakings. These interactions are patterned affairs which provide the context for his own attitude and behavior, and which he, in turn, affects as an active participant. No theory of individual behavior can be adequate until these complex interactions are conceptualized, and until research operations appropriate to their study are developed and used in actual practice.


Bell, W. The utility of the Shevky typology for the design of urban sub-area field studies. Journal of Social Psychology, 1958, 47, 71-83.

———   . Urban neighborhoods and individual behavior, in M. Sherif & C. W. Sherif, eds., Problems of youth: Transition to adulthood in a changing world, pp. 235264. Chicago: Aldine, 1965.

Campbell, J. D. Studies in attitude development: The development of health orientations, in C. W. Sherif & M. Sherif, eds., Attitude, ego-involvement and change, pp. 7-25. New York: Wiley, 1967.

De Soto, C. B., & J. J. Bosley. The cognitive structure of a social structure. Journal of Abnormal and Social Psychology, 1962, 64, 303-307.

Harcum, E. R. Parallel functions of serial learning and tachistoscopic pattern perception. Psychological Review, 1967, 74, 51-62.

Harvey, O. J., & M. Sherif. Level of aspiration as a case of judgmental activity in which ego-involvements operate as factors. Sociometry, 1951, 14, 121-147.

Koslin, B., R. N. Haarlow, M. Karlins, & R. Pargament. Predicting group status from members' cognitions. Sociometry, 1968, 31, 64-75.

MacNeil, M. K. Power of status in norm formation under differing conditions of group solidarity. Ph.D. thesis, University of Oklahoma, 1967.

Pollis, N. P. Relative stability of scales formed in individual, togetherness, and group situations. Ph.D. thesis, University of Oklahoma, 1964.

———. Relative stability of scales formed in individual, togetherness, and group situations. British Journal of Social and Clinical Psychology, 1967, 6, 249-255.

Prado, W. M. Appraisal of performance as a function of the relative ego-involvement of children and adolescents. Ph.D. thesis, University of Oklahoma, 1958.

Sherif, C. W. Variations in judgment as a function of ego-involvements. Paper presented at the meeting of the Eastern Psychological Association, Atlantic City, April 1947.

Sherif, M. In common predicament: Social psychology of inter-group conflict and cooperation. Boston: Houghton-Mifflin, 1966.

Social interaction: Process and products. Chicago: Aldine, 1967.

O. J. Harvey, B. J. White, W. R. Hood, & C. W. Sherif. Intergroup conflict and cooperation: The Robbers Cave experiment. Norman: University of Oklahoma Book Exchange, 1961.

Sherif, M., & C. W. Sherif. Groups in harmony and tension. New York: Harper, 1953; Octagon, 1966.

———. Reference groups: Conformity and deviation of adolescents. New York: Harper, 1964.

Sherif, M., B. J. White, & O. J. Harvey. Status relations in experimentally produced groups. American Journal of Sociology, 1954, 60,370-379.

Shevky, E., & W. Bell. Social area analysis. Stanford, Calif.: Stanford University Press, 1955.


  1. NOTE: The research program reported in this paper was directed by Muzafer Sherif with support from the National Science Foundation, the Rockefeller Foundation, and in its initial stages, the Hogg Foundation for Mental Health. Carolyn W. Sherif collaborated in the research from its inception and had the primary responsibility in preparation of this paper.



Valid HTML 4.01 Strict Valid CSS2