Sei sulla pagina 1di 11

Journal of Personality and Social Piyeholtty

1968, Vol. 10, No. 3, 279-289

DETERMINANTS OF SOCIAL DISTANCE AMONG EAST


AFRICAN TRIBAL GROUPS1
MARILYNN B. BREWER
Northwestern University
Responses to 4 social distance scale items, obtained in interviews with 50
members of each of 30 tribes in Kenya, Uganda, and Tanzania were analyzed
for scalability and sources of variation. According to propositions derived from
several sociopsychological theories of intergroup relations, social distance ratings
of outgroups were predicted to vary with perceived similarity to the ingroup,
physical distance from the ingroup, and educational-economic advancement of
the outgroup. In addition, social distance was expected to be closely related to
ingroup members' liking for and familiarity with the outgroup. Guttman scale
analyses of the social distance item responses revealed that the items were
consistently highly scalable, with a high degree of commonality in the rank
order of social distance steps across the 30 groups. Internal consistency of the
4 items and respondent reliability were also consistently high. Analysis of
variance and correlational analyses of the effects of similarity, distance, and
outgroup advancement, and their interactions, on social distance ratings produced highly consistent results. Social distance was found to vary most strongly
with perceived similarity (the more similar the outgroup, the lower the social
distance), next with distance (the nearer the outgroup, the lower the social
distance), and, to some extent, with advancement (the more advanced the
outgroup, the lower the social distance). There was also evidence of an interaction between similarity and advancement as a determinant of social distance.

Since Bogardus (1925) published his social


distance scale, variations of the original scale
have been used to measure degrees of intimacy
in interaction permitted among various social
groups all over the world (e.g., Catapusan,
19S4; Dawson, 196S; Dodd, 1935; Hunt,
1956; Lambert, 1952; Nataraj, 1965; Pettigrew, 1960; Pro thro & Melikian, 1952; Sinha
& Upadhyaya, 1962; van den Berghe, 1962;
Vaughan, 1962; Zaidi, 1967). Most studies,
however, have been primarily descriptive of
the relations existing between particular social
units. Very little research has been done involving systematic variation of intergroup differences to determine the relative effect of
specific factorsoutgroup characteristics and
the nature of ingroup-outgroup relationson
!This article is based on a doctoral dissertation
submitted to the Graduate School of Northwestern
University, 1968. The author wishes to express her
appreciation to the members of her dissertation committee, particularly to Donald T. Campbell, advisor
and chairman. The research reported in this study
was supported by a grant from the Carnegie Corporation of New York awarded to Northwestern University for the Cooperative Cross-Cultural Study of
Ethnocentrism, under the direction of Donald T.
Campbell and Robert A. LeVine.

the degree of social distance at which outgroups are held. Some systematic research has
been conducted by Triandis and his associates (Triandis & Davis, 1965; Triandis &
Triandis, 1960, 1962) in the United States
and elsewhere. Triandis' research paradigm
involves presenting respondents with descriptions of hypothetical stimulus persons, varying in similarity to the respondent on such
factors as race, nationality, religious belief,
sex, and occupation. Through factorial combinations of these variables, Triandis is able
to determine the relative contribution of each
to variation in social distance ratings of the
stimulus persons. Cross-cultural comparisons
have been made of the relative importance of
various factors. For example, Triandis and
Triandis (1965) reported that race is the most
important factor contributing to variation in
social distance in the United States, while in
Greece religion is most important; in Germany occupation is most important; and in
Japan race and occupation are the major
factors.
The importance of race as a determinant
of social distance in the United States has
been disputed by Rokeach, who maintains

279

280

MARILYNN B. BREWER

that the predominance of this factor in


Triandis' results is an artifact of his methodology. Rokeach's theory of cognitive organization (Rokeach, 1960) holds that the primary
determinant of social distance is the degree
of perceived similarity of belief between the
respondent and the stimulus person. Two
studies (Byrne & Wong, 1962; Stein,
Hardyck, & Smith, 1965) have demonstrated
that in ratings of descriptions of hypothetical
persons by white college students, similar
Negroes are rated more favorably than dissimilar whites. However, if given no information about beliefs, students' ratings of Negro
stimulus persons correlate highly with ratings
of dissimilar whites, suggesting that, in the
absence of specific information, belief-dissimilarity between Negroes and whites is assumed.
A later study by Triandis and Davis (1965)
suggests that the relative contribution of race
and belief to social distance ratings depends
on the nature of the scale items, race being a
more important determinant of responses to
items involving highly intimate behaviors
(e.g., dating, marriage) than of responses to
less intimate items.
Rokeach's theory of the relationship between belief similarity and social distance
accords with the predictions of balance theories of interpersonal relations (Heider, 1946;
Newcomb, 1961). The theoretical basis for
other predictions regarding the characteristics
of an outgroup, or the types of relations beingroup and outgroup, that will lead to the
outgroup's being held at high- or low-social
distance is provided in a monograph by
Campbell and LeVine (1965). Extensions of
major sociological and psychological theories
of intergroup relations lead to the following
propositions which provide the focus of analysis for the present study.
As already mentioned, Rokeach's theory of
cognitive organization and structural balance
theories predict a relationship between social
distance and perceived similarity between ingroup and outgroup: (a) Outgroups perceived
as similar to an ingroup will be held at less
social distance than outgroups perceived as
dissimilar.
Realistic group-conflict theory predicts that
nearer outgroups will be the targets of intergroup hostility since they offer the greatest

threat in competition for scarce resources.


Thus, the following prediction is generated
regarding the relationship between physical
distance and social distance: (&i) Adjacent
outgroups will be held at greater social distance than nonadjacent outgroups. However,
in many cases, adjacency and cultural similarity occur simultaneously, which places the
realistic conflict theory in contradiction to the
prediction derived from balance theory. In
addition, social-structural theories of conflict
emphasize that, at least for groups with a
pyramidal-segmentary type of group loyalty
structure, adjacent groups are most frequently
allies in common defense against remote
groups, providing further theoretical basis for
the conflicting prediction: (b2) Adjacent outgroups will be held at less social distance than
nonadjacent outgroups.
As a compromise between balance theory
and realistic group conflict, an interaction between perceived similarity and physical distance as determining social distance may be
predicted: (ab) Among outgroups perceived
as similar to an ingroup, adjacent groups will
be held at less social distance than nonadjacent groups, but among dissimilar outgroups, adjacent groups will be held at greater
social distance than nonadjacent groups.
Another source of variation in social distance ratings is suggested by frustrationaggression theory which predicts that hostility
will be directed against outgroups perceived
as most frustrating to the ingroup, such as
stronger, more aggressive, and more prestigeful outgroups. Assuming that the relative
educational-economic advancement of outgroups is a source of frustration to ingroup
members, the following prediction is generated: (ci) High-advanced outgroups will be
held at greater social distance than lowadvanced outgroups. However, reference-group
theory maintains that social groups will admire and emulate successful outgroups, leading to the opposite prediction: (c^) Highadvanced outgroups will be held at less social
distance than low-advanced outgroups.
Perhaps whether outgroup advancement is
a source of frustration or admiration for ingroups depends on whether the advanced
group is perceived as similar or dissimilar.
Thus, an interaction between perceived simi-

SOCIAL DISTANCE AMONG EAST AFRICAN TRIBAL GROUPS


larity and advancement as a determinant of social distance may be predicted: (ac) Among
outgroups perceived as similar to the ingroup,
high-advanced groups will be held at less
social distance than low-advanced groups, but
among dissimilar outgroups, high-advanced
groups will be held at greater social distance
than low-advanced groups. Another modification of Hypothesis Ci is provided by
frustration-aggression theory itself, which predicts that overt expression of hostility will be
avoided against groups in a good position to
retaliate. Thus, the expression of hostile
evaluations against advanced outgroups may
depend on the physical distance between ingroup and outgroup, leading to the following
interaction prediction: (be) High-advanced
outgroups will be held at less social distance
if they are adjacent to the ingroup than if
they are nonadjacent.
The nature of social distance, as a measure
of intimacy of intergroup relations and a
reflection of favorable or unfavorable attitudes
in intergroup perception, leads to the prediction of two other correlates of social distance ratings: (d) There will be a positive
corrrelation between social "nearness" and
liking for outgroups; and (e) There will be
a positive correlation between social "nearness" and familiarity with outgroup members.
The purpose of the present study was to
determine the predictive validity of the preceding propositions in the relations among
tribal groups of East Africa.
METHOD
Data Collection
The data for the present study came from a survey
conducted among 30 tribes of East Africa by Marco
Surveys, Ltd., under the direction of Gordon Wilson
in Kenya. The survey interview and sample plan
were prepared by Robert A. LeVine and Gordon
Wilson, using a field manual previously prepared by
LeVine and Campbell (1965). Fifty respondents
were interviewed in each of 10 tribes in Kenya,
Uganda, and Tanzania.2 Respondents represented all
2

The tribes included in the survey from Kenya:


Kikuyu, Meru, Embu, Kamba, Kisii, Kipsigis, Masai,
Luo, Luhya, and Nandi; from Uganda: Ganda,
Banyoro, Batoro, Ankole, Basoga, Bagisu, Teso,
Acholi, Lango, and Karamojong; from Tanzania:
Sukuma, Nyamweri, Luo, Gogo, Chagga, Pare,
Sambaa, Meru, Arusha, and Masai. Some comment
on the use of the term "tribe" to refer to these social

281

levels of age, education, sex, occupation, and religious


affiliation, but with high proportions of uppereducated young men (secondary-school education).
All interviews were conducted individually, orally,
and usually in the tribal native language (but some in
English, depending on the respondent's preference).
The interviewer recorded the responses in English
and later coded them in a numerical coding system.
The tribes were chosen to represent roughly comparaable social units for that area of the world.
The interview schedule contained three major sections each involving the relations between a respondent's own tribe (ingroup) and 13 outgroups9
tribes from the same country and 4 from either of
the other two countries included in the survey. The
first section included four social distance questions
and three questions involving familiarity of the
respondent with outgroup members, asked about each
of the 13 outgroups. The second section required the
respondent to name "the most important good thing"
and "the most important bad thing" about each outgroup. The third section of the interview involved
a series of 48 traits, both positive and negative, to
which the respondent was to name the tribe, among
the 14 (ingroup and outgroups) available, for which
each trait was most characteristic. Four additional
questions asked the respondent to name the outgroup
most liked by, most disliked by, most similar to,
and least similar to his own tribe. The purpose of
the entire interview was to establish patterns of
intergroup relations and perceptual stereotypes among
the 30 tribes included in the survey.
The data for the present study were provided by
the four social distance questions in Section I of
the interview schedule. The social distance items
were worded in English as follows:
1. Would you willingly agree to work with a
?
2. Would you willingly agree to have a
as a neighbor in your house?
3. Would you willingly agree to share a meal with
a
?
4. Would you willingly agree to become related to
a
by marriage?
Responses to each of the social distance questions
were scored 0 for a "no" response, 1 for "don't
know," and 2 for "yes." Thus, the higher the score,
the lower the social distance toward the outgroup.
A total social distance score for each respondent for
each outgroup was obtained by summing his responses to the four items. Possible total scores
ranged from 0 to 8, with the highest score repreunits is called for in view of the controversy among
American anthropologists on the use of the term.
Discussion of the various points of view on this
issue is available in the proceedings of a recent
meeting of the American Ethnological Society (Helm,
1968). For purposes of this paper it was felt that
the term tribe, however vague its reference, was
preferable to the more lengthy and awkward
"traditional ethnic-linguistic unit."

282

MAEILYNN B. BREWER

senting least social distance. For each ingroup, the


average social distance rating of each outgroup was
obtained by computing the mean rating of the SO
respondents toward that outgroup.

Scale Analyses
To determine the scalability and internal consistency of the four items comprising the social distance
scale, responses from each tribe were subjected to
Guttman Scalogram Analysis and item-intercorrelation analysis. The Guttman analysis involved obtaining a reproducibility coefficient (Rep) on dichotomous scores (Responses 0 and 1 = 0, and Response
2 = 1), with the items ordered according to item
marginals. Item intercorrelations were computed with
the Kuder-Richardson Formula 20. Both analyses
were made separately for each of the 30 tribes, with
an N of 6SO for each (responses of SO respondents
for 13 outgroups). In this way, the quality of the
data in meeting the assumptions of unidimensionality and scalability could be compared across groups
of respondents. In addition, the reliability of the
respondent ratings were assessed within each tribe
by computing a correlation between the social distance responses of half of the interviewees (randomly
chosen) and those of the other half, across the 13
outgroups.

Analyses of Sources of Variation


In order to determine the relative contribution
of the three major variables (perceived similarity,
physical distance, and educational-economic advancement) to variation in social distance ratings of outgroups, responses of each ingroup were measured
toward eight outgroups selected to represent a factorial combination of the three variables, with two
levels of each. The first two variables, distance and
similarity, were relational, or dyadic variables, defined by characteristics of both the ingroup and the
outgroup. The two levels of distance were adjacent
and nonadjacent, adjacency being defined as the
sharing of a common border by ingroup and outgroup.
Perceived similarity was denned by the frequency
of mention of the outgroup by ingroup members in
response to the questions "which tribe is most similar
to your own tribe" and "which tribe is least similar
to your own tribe." Each outgroup was classified as
"similar" or dissimilar" according to the net frequency of mentions as "most similar" minus those
as "least similar."
The third variable, economic-educational advancement, was a monadic variable, involving a characteristic of the outgroup independent of the ingroup.
Each tribe in the survey was classified as high or
low on this variable according to frequency of mentions, across all responding groups, on trait questions such as "intelligent," "stupid," "backward,"
"progressive," "keen to advance," "conservative,"
"wealthy," and "poor." (Assignment of groups to
these categories was done with the cooperation of

Robert A. LeVine, an anthropologist familiar with


the tribal groups of East Africa.)
By obtaining social distance ratings of eight outgroups representing all combinations of the two
levels of similarity, distance, and advancement, the
relative contribution of each of the three variables
and their interactions to variation in ratings could
be assessed by analysis of variance. However, the
major problem with the design was that the outgroups about which respondents in each ingroup
were questioned had not been originally selected to
represent a factorial combination of the three variables chosen for study here. Therefore, the available
set of outgroups had to be "forced" into the cells
required for a 2 X 2 X 2 design. For instance, in order
to complete the design for a particular ingroup, two
moderately advanced outgroups may have to be
differentiated as relatively "high" or "low" on the
educational-economic advancement dimension.
The necessity of forcing some cells was not expected to affect seriously the main effects obtained
in the analysis of variance since the levels of each
variable would remain intact. If anything, a slight
negative bias was introduced against finding significant main effects since the two levels of each variable were less sharply differentiated than if the
selection of outgroups could have been more controlled. However, if the forcing procedure differentially affected certain cells of the factorial design,
the interaction effects could have been affected by
the introduction of a positive bias. For instance,
if it were more difficult to obtain "pure" examples
of nonadjacent, perceived similar outgroups than
adjacent, similar outgroups, then a finding of a
significant interaction between distance and perceived
similarity could be an artifact of the design rather
than a true interaction.
Because of this methodological difficulty inherent
in the analysis of variance design, a separate assessment of the effects of the three major variables was
made using correlational techniques and including all
13 outgroups rated by each of the 30 ingroups.
For this purpose, each outgroup was given a score
on each of the three independent variables. Perceived
similarity was defined as the frequency of mentions
of the outgroup by the ingroup as "most similar"
less the frequency of mentions as "least similar."
Distance was measured on a 4-point scale with 4
assigned to adjacent groups, 3 to outgroups with one
tribe between their territory and that of the ingroup,
2 to outgroups with two tribes between, and 1 to
more remote groups. (Determination of distances
was based on tribal maps of East Africa provided
in Good, 1966.) Outgroup advancement was also
rated on a 4-point scale on the same basis as described for the assignment of groups to levels of
advancement in the factorial design. Of the 30 tribes
included in this study, 3 were rated as 4 (high
advanced), 9 as 3 (moderately high advanced), 14
as 2 (moderately low advanced), and 4 as 1 (low
advanced).
In addition to obtaining the correlation between
mean social distance ratings of each outgroup and

SOCIAL DISTANCE AMONG EAST AFRICAN TRIBAL GROUPS


its scores on each of the three variables, the effect
of interactions of the variables was assessed by a
method used by econometricians (Draper & Smith,
1966; Goldberger, 1964) involving correlations between social distance and the cross products of the
independent variable scores, partialing out the main
effect correlations.
The relationship between social distance and acquaintance with and liking for outgroups was also
obtained through correlation. Acquaintance was measured for each ingroup respondent relative to each
of the 13 outgroups in terms of living among members of the outgroup and speaking their language.
For each respondent, an acquaintance score was
obtained for each outgroup, ranging from 0 (never
lived among them nor spoke their language) to S
(lived among them for three years or more and
speaks their language). The mean acquaintance with
each outgroup for the SO respondents of each ingroup was correlated with the mean social distance
rating of that outgroup. Similarly, an index of liking
for each outgroup, obtained by subtracting the frequency of mentions of the outgroup in response to
the question, "which tribe is most disliked by your
own tribe," from the frequency of mentions in response to "which tribe is most liked by your own
tribe," was correlated with mean social distance.

RESULTS
Scale Analyses
The results of the various analyses of unidimensionality and reliability of the social
distance scale items for each of the 30 tribes
in this study were highly consistent.8 The
Rep values, computed for each tribe, were
uniformly high, ranging from .91 to .99, indicating that the scalability of the items was
good. However, of more interest is the order
of scaled items obtained for each of the tribes
(based on item marginals, or proportion of
yes responses), since these provide a test of
the implicit assumption that there is some
commonality in the rank ordering of social
distance steps. The items as listed in the
Method section were presumably ordered from
least intimate (1) to most intimate (4), but
this rank ordering was obtained for only 6 of
the 30 tribes in this study. For 19 out of the
30, the obtained order was 3-1-2-4. (Considering all possible item orders as having an
equal chance of occurring, the binomial probability of this one order occurring so frequently by chance is less than .00000001.) To
test whether the obtained deviations from this
3
Copies of an extended table reporting scaleanalysis results for each tribe are available from the
author.

283

item order would be expected by chance, a


multivariate analysis of variance was performed on the item-response patterns for the
30 tribes, dividing the respondents of each
tribe randomly in half to get an estimate of
within-tribe variation of responses. No significant difference was found. Thus, there is
strong indication that among the groups surveyed there was a regular ordering of social
distance steps as follows, from least intimate
to most intimate:
1. Willingness to share a meal with outgrouper;
2. Willingness to work with outgrouper;
3. Willingness to have outgrouper as a
neighbor; and
4. Willingness to become related to outgrouper by marriage.
That sharing a meal is considered less intimate than working together or living in the
same neighborhood is not surprising if it is
realized that, traditionally, visiting strangers
were treated to meals and that such hospitality was an important factor in a tribe's
prestige. Thus, within this cultural area, the
consistency of scale steps is attributable to
shared traditions regarding intergroup social
contact.
The lack of significant differences among
tribes can also be attributed, in part, to the
fact that differentiation among items (in terms
of proportion of yes responses) was generally
quite small. Examination of individual response patterns revealed that this was not
due to variation in respondents' interpretation of the four items but rather to a tendency
to give "all or nothing" ratings to outgroups. Thus, although all the obtained Rep
values were greater than expected by chance
(based on item marginals) their size was
largely due to high proportions of "all yes"
or "all no" response patterns. Some tribes,
particularly the Kenya and Tanzania Masai,
were characterized by huge proportions of
"all no" response patterns, as indicated by
very low and undifferentiated marginals.
Other tribes, like the Kisii, Kipsigis, and
Arusha, made high proportions of "all yes"
responses, while the rest of the groups were
more differentiating among outgroups. Only
five tribes (Kamba, Meru, and Luo of Kenya,

284

MAKILYNN B. BREWER
TABLE 1
ANALYSIS OF VARIANCE OF MEAN SOCIAL
DISTANCE RATINGS
Source

if

MS

Tribes (T)
Similarity (S)
Distance (D)
Advancement (A)
SXD
SX A
DX A
SXDXA
TXA
Pooled error
Total

29
1
1
1
1
1
1
1
29
174
239

14.28
149.47
45.82
17.03
0.94
4.77
0.00
0.28
1.33
1.45

9.85**
103.08**
31.60**
12.80**
0.65
3.29*
0.00
0.19
0.92

* t < .10.

**p < .01.

and Acholi and Sukuma) exhibited differences


of as much as .40 between any item marginals,
indicating some differential responding to
different scale items.
The internal reliability of the social distance scale, as measured by the KuderRichardson Formula 20, was consistently high
across the 30 tribes, especially for a scale of
only four items. Applying the Horst correction for uneven item marginals produced even
higher reliabilities, ranging from .55 to .98
for 29 of the tribes. (Lack of variation among
the responses of the Kipsigis respondents
accounts for the occurrence of one .00 correlation.) Across tribes, the split-half reliability
of respondents was .99 (N = 30), and it was
also consistently high within tribes. However,
at least part of this intratribe consistency of
responses may be accounted for by the fact
that only one interviewer handled all the
interviews within one tribe.
Analyses of Sources of Variation
The results of the analysis of variance of
the mean social distance rating by each ingroup of the eight selected outgroups are
reported in Table 1. A repeated-measures
model of analysis of variance was used in
which it was assumed that no interactions
with tribes existed except the Tribe X Advancement interaction, which was of some
interest. The other interactions were considered error variance and were combined for
a pooled error term.
All four main effects were significant at the
.01 level, but there were clear differences in

the strength of the effects. The significant effect of tribes indicates that, as expected, there
were differences among the tribal groups in
overall social distance ratings given to outgroups generally. The direction and strength of
the other main effects, across the 30 tribes,
are indicated by the mean values reported in
Table 2. The effect of similarity was clearly
the strongest and in the direction predicted
by Hypothesis a. The next strongest effect
was physical distance, such that adjacent
outgroups were held at less social distance
than nonadjacent outgroups, supporting Hypothesis Z>2 and contrary to Hypothesis b\.
The weakest effect was that of outgroup advancement, which was in the direction predicted by Hypothesis c2 and opposite to that
predicted by Hypothesis c\. The lack of any
interaction between advancement and tribes
indicates that high-advanced groups were held
at less social distance than low-advanced outgroups irrespective of the degree of ingroup
advancement. Distance between ingroup and
outgroup also did not influence the positive
effect of outgroup advancement, providing no
support for Hypothesis be.
Hypothesis ab, predicting an interaction between similarity and distance, received no
support. The positive effect of distance was
the same for similar and dissimilar outgroups.
However, the interaction between similarity
and advancement almost reached significance.
Examination of the means in Table 3 reveals
that the nature of the interaction was not as
predicted in Hypothesis ac, but rather was
such that for similar outgroups there was little
difference in the social distance ratings of
high- and low-advanced groups, but among
dissimilar outgroups high-advanced groups
were held at less social distance than lowadvanced groups.
TABLE 2
MAIN EFFECT MEANS
Variable
Rating
Similarity

High
Low
Adjacent
Nonadjacent

5.2
3.6

Note.N = 120 per mean.

Distance

4.8
3.9

Advancement

4.7
4.1

285

SOCIAL DISTANCE AMONG EAST AFRICAN TRIBAL GROUPS


TABLE 3
SIMILARITY X ADVANCEMENT INTERACTION
Similarity
Advancement

High
Low

High

Low

5.3

4.0
3.2

5.1

Note.N 60 per cell mean.

Since the interaction and relative strength


of main effects obtained in the analysis of
variance may have been affected by the
methodological difficulties of forcing outgroups into a factorial design, support for
these findings was sought using correlational
analysis. In addition to providing another
method of testing the major hypotheses, the
correlational analysis had the advantage of
using more dataratings of 13 outgroups for
each ingroup instead of 8, and several levels
of each independent variable instead of only
two. Obtained correlations between the major
variables and their cross-products and mean
social distance are reported in Table 4 along
with those of the minor variables, liking and
familiarity. Since correlations computed across
tribes would not control for different levels of
responding to the social distance scale by
different ingroups, the variation in responses
within each group (N = 13 each) was related
to the selected variables and the average of
the 30 intratribe correlations was computed.
These average correlations were partialed to
estimate the effect of each variable holding
some of the others constant.
The results of the correlational analysis are
strikingly similar to those of the analysis of
variance. All three main variables were positively correlated with social distance scores
and the rankings of the main effects in order
of degree of relationship was the same as that
of the mean squares in the analysis of variance results. The effect of similarity was
strongest and clearly positively linear. The
correlation between physical distance and
social distance could have been inflated because of the expected positive correlation
between nearness and similarity (r = .28),
but after partialing out the effect of similarity,
the distance correlation remained high and
positive. The correlation between social dis-

tance and outgroup advancement, however,


was suppressed because the selection of outgroups to be rated produced a slight negative
correlation (r= .16) between advancement
and distance. This was true because the most
remote outgroups selected (those from other
countries) were usually high-advanced groups.
Controlling for this effect, the advancementsocial distance correlation was raised from
.26 to .41.
The interaction correlations reported in
Table 4, with the main effects partialed out,
represent the relationship between variation in
the products of the main variables and residual social distance (obtained social distance
scores minus predicted values based on the
main effects), or variation in social distance
not accounted for by linear combinations of
the three major variables. Consistent with the
results obtained from the analysis of variance, the correlations provide no evidence for
interactions between similarity and distance,
distance and advancement, or similarity, distance, and advancement as determinants of
social distance. The product of similarity and
advancement, however, did correlate significantly with social distance in the negative
direction. This finding was compared with the
interaction tendency obtained from the analysis of variance. The mean values reported in
Table 3 correspond to a negative interaction
effect in that the high-high and low-low cell
TABLE 4
CORRELATIONS BETWEEN MEAN SOCIAL DISTANCE
AND OTHER VARIABLES
Variable

Perceived similarity (S)


Physical distance (D)
Outgroup advancement (A)
SXD
SXA
DXA

SXDXA
Liking
Familiarity

Average intratribe
correlation"

.66*
.54* (.50 with S
partialed)
.26* (.41 with D
partialed)
.00 (with S and
D partialed)
-.53* (with S and
A partialed)
.06 (with D and
A partialed)
-.04 (withS, D,
and A partialed)
.71*
.77*

N "30. Significance levels for these values were determined by a t test of the hypothesis that the mean of the 30
correlations (r.) was equal to .00.
*t> <.01.

286

MARILYNN B. BREWEK
TABLE S
AVERAGE INTERCORRELATIONS AMONG FOUR
RATINGS OF OUTGROUPS
lutercorrelation
Var'-ihl '

1. Social distance
2. Perceived similarity
3. Liking
4. Familiarity

.66
.71

.77

.69
.54

.48

means are lower than would be expected by


the size and direction of the similarity and
advancement main effects alone, whereas the
high-low and low-high cell means are higher
than expected. Thus, the correlation results
support the analysis of variance findings, although the relative size of the interaction
correlation is much higher than that of the
interaction mean square.
The correlations reported in Table 4 also
include those between social distance and
liking for and familiarity with the outgroup.
Both of these indexes were strongly related
to mean social distance ratings of outgroups,
providing support for the concept of social
distance as an index of favorability of attitudes toward and degree of interaction with
outgroups.
The consistency of results obtained from
both analyses of sources of variation in
social distance ratings prompted further
consideration of each of the related variables.
Similarity and Social Distance
The balance models of cognitive organization predict that, in order to maintain consistency among related cognitions, individuals
will like others whom they perceive to be similar to themselves and will dislike those whom
they perceive to be dissimilar. Rokeach's
(I960) theory specifies that social distance
between individuals and members of other
social groups varies with perceived similarity.
Extending his argument to the level of intergroup relations, it was predicted that the average social distance ratings of outgroups would
be a function of the extent to which ingroup
respondents perceived the outgroup as similar
to their own tribe. This prediction was
strongly supported by the analysis of variance results and the obtained positive cor-

relation (r = .66) between similarity and


social distance scores. Furthermore, the analysis of variance results indicated that the effect
of similarity was as strong for distant outgroups as it was for adjacent outgroups.
Examination of the intratribe correlations
revealed that the relation between similarity
and social distance was consistently positive
across the 30 tribes, with surprisingly low
variation considering that each correlation
was based on an N of 13 (the standard deviation of the 30 correlations was .07). However,
the effect of similarity on social distance cannot be isolated from the effects of two closely
related variablesliking and familiarity. The
average intercorrelations among these four
variables are reported in Table 5. A simple,
direct method of extracting a single factor
(Harman, 1960, p. 122) was applied to this
matrix and resulted in high positive loadings
for each variable (ranging from .71 to .95)
with very small residual matrix, indicating
unidimensionality. Thus, these data provide
strong evidence that social distance, perceived
similarity, liking, and familiarity are cosymptoms of an intimacy-friendship syndrome in
intergroup relations.
Physical Distance and Social Distance
Contradictory propositions about the effects
of physical distance on social distance were
derived from balance theory and from realistic
group-conflict theory. To the extent that
adjacency between tribes is related to their
degree of similarity, balance theory leads to
the prediction that nearer outgroups will be
held in less social distance than more remote
outgroups. On the other hand, realistic groupconflict theory, stressing the opportunity for
competition over scarce resources between
neighboring tribes, leads to the prediction that
nearer tribes will be held in greater social
distance than more remote groups. Only the
former prediction was confirmed by the results of the analysis of variance and the obtained positive correlation (r = .54) beween
nearness and social distance scores. In addition, the analysis of variance results indicated
that adjacent groups were held at lower social
distance than remote groups whether perceived as similar or dissimilar to the ingroup.
Thus, the interaction predicted in Hypothesis

SOCIAL DISTANCE AMONG EAST AFRICAN TRIBAL GROUPS


ab, derived as a compromise between the
conflicting predictions of balance and realistic
group conflict, was not supported.
Another compromise between the conflicting
predictions was considered by examining the
relationship between social distance and physical distance for curvilinearity. Even if nearer
groups are generally better liked than remote
groups, it could be that the effect of threat
from immediate neighbors would make them
less liked than close but not adjacent outgroups. Such a trend could not be observed
from the analysis of variance results, where
only two levels of physical distance were considered, but it was possible to compute average ratings for each ingroup of outgroups at
each of the four levels of distance considered
in the correlational analysis. The mean
values, across the 30 tribes, of these average
ratings are plotted in Figure 1 and show no
evidence of curvilinearity; immediately adjacent outgroups were held at less social distance than outgroups at any other level of
distance.
Although the cross-tribe results indicate a
positive relationship between physical distance and social distance, examination of the
varation among intratribe correlations revealed that the realistic group-conflict prediction was not entirely without support. The
mean correlation was significantly positive,
but the variation was greater than that for
similarity (SD = .08) and the correlation for
one tribe, the Kisii of Kenya, was highly
negative (r = .61). The Kisii are geographically separated from the other Kenya Bantu
tribes to which they are most similar, and
of their three immediate neighboring tribes,
two (Kipsigis and Masai) are traditional
enemies of the Kisii, with a record of quite
recent hostilities, and the other (Luo) is
looked down upon by the Kisii because they
do not practice the Kisii custom of circumcision. Thus, there is some evidence that extent
of military or social conflict between adjacent
groups can seriously affect the relationship
between physical nearness and social distance.
Out group Advancement and Social Distance
As with the effects of physical distance,
contradictory predictions can be derived for
the effects of outgroup advancement on social

b.
O

287

6.0

5.0

4.0

in
3.0

o
o
in

2.0

DISTANCE BETWEEN INGROUP


AND OUTGROUP

FIG. 1. Social distance as a function of


physical distance.

distance ratings, depending on whether one


considers the advanced group as an object of
admiration or as a source of frustration for
the less advanced groups. The results of the
analysis of variance and correlation in this
study support only the former orientation in
that high-advanced outgroups were held at
less social distance than low-advanced outgroups. The fact that this relationship held
for all ingroups (i.e., whether high or low advanced) can be considered support for reference-group (emulation-admiration) theory if
one assumes that the greater liking for highadvanced outgroups was a function of admiration on the part of the low-advanced ingroups
and of disrespect for low-advanced outgroups
on the part of high-advanced ingroups. This
interpretation is supported by the evidence
obtained for an interaction between similarity
and advancement as sources of social distance. For outgroups perceived as similar to
the ingroup, little difference in the social distance ratings of high- or low-advanced groups
was obtained, but for dissimilar outgroups
(i.e., those whose differential advancement
was recognized and considered) high-advanced
groups were held at less social distance than
low-advanced groups.

288

MARILYNN B. BREWER

a.

i
s
O

* all outgroupt
* within-country outgroupt

6.0

b.
O

Ioc "
111
u
z

4.0

3.0

o
I/I

2.0

ADVANCEMENT OF OUTGROUP

FIG. 2. Social distance as a function of


outgroup advancement.

Again because of the contradictory predictions involved, the relationship between advancement and social distance was examined
for curvilinearity. If high-advanced groups
were objects of mixed reaction, a combination
of admiration and resentment, from lowadvanced groups, then slightly less advanced
groups who were still admired but less resented may be held at less social distance. The
mean ratings of outgroups at four levels of
educational-economic advancement are plotted
in Figure 2. The relationship between social
distance and outgroup advancement appears
to be curvilinear when ratings of all outgroups
are included. However, it has already been
noted that very remote tribes were overrepresented among the high-advanced outgroups.
When these groups are omitted, and only
ratings of outgroups within the ingroup's
own country are considered, the relationship
between social distance and advancement
becomes more linear, as indicated by the
second curve in Figure 2. Even within the
same country, moderately high-advanced outgroups appear to be rated as high as very
advanced groups, but this probably reflects
the fact that objectively there was less difference between groups at these two levels of
advancement than between those at any other
levels.

DISCUSSION
The purpose of the present study was twofold: to validate the concept of social distance, as a ranking of degrees of permitted
social intimacy, and to determine some of the
covariates of social distance ratings among 30
tribes of East Africa. Regarding the first
purpose, the four social distance items included in the survey interview were found to
be consistently scalable across the 30 tribes,
with a high degree of commonality in the rank
order of social distance steps. In addition,
differential social distance ratings of outgroups was found to be highly consistent
among the SO interviewees of each ingroup.
Results of both intertribe analysis of variance and intratribe correlational analyses indicated that social distance toward outgroups
increases with increases in perceived dissimilarity to the ingroup and physical distance
from the ingroup, and decreases with increased educational-economic advancement of
the outgroup. Both analyses also indicated an
interaction between the effects of perceived
similarity and outgroup advancement on social distance such that the difference in ratings
of high- and low-advanced outgroups was less
for outgroups perceived as similar to the ingroup than for dissimilar groups. In general,
the results obtained support propositions
derived from balance models and referencegroup-theory approaches to intergroup relations and fail to support those derived from
realistic group-conflict or frustration-aggression theories. The model followed here, of
testing competing hypotheses derived from
different sociopsychological theories of intergroup relations, was obtained from Campbell
and LeVine (1965), and the data obtained
from this study will contribute to a larger
body of cross-cultural material being collected
to make comparisons among many such theories. Thus, although the present findings regarding intertribe social distance differentially
support some theories over others, many other
aspects of intergroup relations, on which the
theories provide complementary or contradictory propositions, will have to be examined before the relative predictive validity
of the various theoretical positions can be
established.

SOCIAL DISTANCE AMONG EAST AFRICAN TRIBAL GROUPS

289

Limitations on the generalizability of the HARMAN, H. H. Modern factor analysis. Chicago:


University of Chicago Press, 1960.
findings of this study are imposed by the
HEIDER, F. Attitudes and cognitive organization.
nature of the social units being considered.
Journal of Psychology, 1946, 21, 107-112.
Relationships among tribal groups in East HELM, J. (Ed.) Essays on the problem of tribe:
Proceedings of the 1967 Annual Spring Meeting of
Africa are in a state of flux; traditional relathe American Ethnological Society. Seattle: Unitions being broadened or superceded by rapid
versity of Washington Press, 1968.
economic and political change. The impor- HUNT,
C. L. Social distance in the Philippines.
tance of economic roles and opportunities for
Sociology and Social Research, 1956, 40, 253-260.
intertribe contact as determinants of inter- LAMBERT, W. E. Comparisons of French and American modes of response to the Bogardus social
group perceptions has been indicated by
distance scale. Social Forces, 1952, 31, 155-164.
studies of stereotype content (Campbell,
LEVINE, R. A., & CAMPBELL, D. T. Ethnocentrism
1967) and is supported here by the strong
field manual. Department of Psychology, Northeffect of familiarity on social distance ratings.
western University, 1965. (Mimeo)
Therefore, the determinants of social distance NATARAJ, P. Social distance within and between
castes and religious groups of college girls. Journal
found among these tribal groups may not
of Social Psychology, 1965, 65, 135-140.
generalize to the relations among other types
NEWCOMB, T. M. The acquaintance process. New
of social groups, particularly political units.
York: Holt, Rinehart & Winston, 1961.
Another limitation of the present study is PETTIGREW, T. F. Social distance attitudes of South
African students. Social Forces, 1960, 38, 246-253.
the fact that all interviews within one tribe
were conducted by the same interviewer, PROTHRO, E. T., & MELIKAN, L. Social distance and
social change in the Near East. Sociology and
which may account for much of the consistSocial Research, 1952, 37, 3-11.
ency of intratribe response patterns. Inter- ROKEACH, M. The open and closed mind. New York:
viewer effects may account for some of the
Basic Books, 1960.
differences found among tribes in overall so- SINHA, A. K. P., & UPADHYAYA, O. P. Eleven ethnic
groups on a social distance scale. Journal of Social
cial distance ratings but could not contribute
Psychology, 1962, 57, 49-54.
to the major effects obtained across tribes STEIN, D. D., HARDYCK, J. A., & SMITH, M. B. Race
and validated by two types of analysis.
and belief: An open and shut case. Journal of
REFERENCES
BOOARDUS, E. S. Measuring social distances. Journal
of Applied Sociology, 1925, 9, 299-308.
BYRNE, D., & WONG, T. J. Racial prejudice, interpersonal attraction, and assumed dissimilarity of
attitudes. Journal of Abnormal and Social Psychology, 1962, 65, 246-2S3.
CAMPBELL, D. T. Stereotypes and the perception of
group differences. American Psychologist, 1967, 22,
817-829.
CAMPBELL, D. T., & LEVINE, R. A. Propositions about
ethnocentrism from social science theories. Department of Psychology, Northwestern University,
1965. (Mimeo)
CATAPUSAN, B. T. Social distances in the Philippines.
Sociology and Social Research, 19S4, 38, 309-312.
DAWSON, J. Race and intergroup relations in Sierra
Leone. Part 2. Race, 1965, 6, 217-231.
DODD, S. C. A social distance test in the Near East.
American Journal of Sociology, 1935, 41, 194-204.
DRAPER, N. R., & SMITH, H. Applied regression analysis. New York: Wiley, 1966.
GOLDBERGER, A. S. Econometric theory. New York:
Wiley, 1964.
GOOD, C. M. Dimensions of East African culture.
East Lansing, Mich.: African Studies Center, 1966.

Personality and Social Psychology, 1965, 1, 281289.


TRIANDIS, H. C., & DAVIS, E. E. Race and belief as
determinants of behavioral intentions. Journal of
Personality and Social Psychology, 1965, 2, 715725.
TRIANDIS, H. C., & TRIANDIS, L. M. Race, social class,
religion, and nationality as determinants of social
distance. Journal of Abnormal and Social Psychology, 1960, 61, 110-118.
TRIANDIS, H. C., & TRIANDIS, L. M. A cross-cultural
study of social distance. Psychological Monographs, 1962, 76(21, Whole No. 540).
TRIANDIS, H. C., & TRIANDIS, L. M. Some studies of
social distance. In I. D. Steiner & M. Fishbein
(Eds.), Current studies in social psychology. New
York: Holt, Rinehart & Winston, 1965.
VAN DEN BEROHE, P. L. Race attitudes in Durban,
South Africa. Journal of Social Psychology, 1962,
57, 55-72.
VAUGHAN, G. M. The social distance attitudes of
New Zealand students toward Maoris and fifteen
other national groups. Journal of Social Psychology, 1962, 57, 85-92.
ZAIDI, S. M. H. A study of social distance as perceived by students of Karachi University. Journal
of Social Psychology, 1967, 71, 197-207.
(Received April 15, 1968)

Potrebbero piacerti anche