Вы находитесь на странице: 1из 24

Multiplexity in Adult Friendships*

LOIS M. VERBRUGGE,

UniversityofMichigan

ABSTRACT "Multiplexity" is the overlap of roles, exchanges, or affiliations in a social relationship. In two cities, adult friendship dyads are examined for multiplexity in three affiliations: kin, neighbor, and coworker. Several factors are proposed to account for multiplex ties: limited opportunities for social contact; preference for the special similarities that kin, neighbors, and coworkers have; and preference for holistic and diffuse friendships. Two structural features are discovered: repetition of multiplexity, and segregation of affiliations. These mean that if multiplexity appears in one friendship, it tends to be repeated in others. But the categories kin, neighbor, and coworker seldom appear in the same friendship. Repetition and segregation are demonstrated by comparing observed friendships with expectations Erom a random-choice model. When social and demographic differentials in multiplexity are examined, they confirm the importance of both opportunities and preferences in motivating multiplex friendships. Consequences of multiplexity for friendship behavior are hypothesized, and one consequence (contact frequency) is analyzed. Neighbor multiplexity increases friendship contact, a reflection of high opportunities for contact among neighbor friends. Kin and coworker multiplexity do not increase friendship contact.

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

When adults develop close friendships, they sometimes select their friends from among kin, neighbors, and coworkers. In social networks research, this overlap of two or more important social relationships is called multiplexity. Although some studies report the frequency of such overlap, no study has seriously examined the structure and importance of multiplexity in adult friendships. Does multiplexity occur throughout a person's friendship network? In other words, if one close friend is a relative, are other close friends also relatives? Are friendships often highly multiplex, for example, kin friends who are coworkers or neighbors as well? What encourages people to develop multiplex friendships, and what consequences does multiplexity have for the stability and energy of friendships? This paper concentrates on the structure of multiplexity and its causes in adult friendships in two urban populations, with less attention to
*This a revised version of a paper presented at the annual meeting of the American Sociological Association, 1976. The author thanks E. O. Laumann for use of the Detroit survey data, both Laumann and F. U. Pappi for the Altneustadt data, and reviewers for their comments.
1979, University of North Carolina Press. 0037-77321791570417470$02.40

1286

Adult Friendships 1 1287


the consequences of multiplexity. In both sites, adults were asked to name their close friends. For this analysis, a friendship is called multiplex if the two people are also kin, or neighbors, or coworkers. Clearly, there are other ways to enumerate multiplexity in adult social ties: one could ask how many of a person's relatives are also friends, or which neighbors are close friends, etc. All such data produce conditional probabilities, e.g., given that two people are friends, what is the chance they are also kin (or coworkers, or neighbors)? This is the only feasible approach to study multiplexity, since one cannot hope to enumerate all social ties in a sample in order to estimate the joint probabilities of two people being kin-and-friend, coworker-and-neighbor, etc. The paper is organized as follows: First, the frequency of multiplex friendships in the two study sites is described. Then, the structure of multiplex ties is examined. Two structural features, called repetition and segregation, are discovered: (1) People who choose a kin as a close friend tend to choose kin for their other close friends. (Similarly for neighbor and coworker friends.) The converse is also true: people who avoid multiplexity in one close friendship tend to avoid it in all close friendships. (2) Few people have friends who overlap in two or three ways (e.g., are both kin and coworker). The factors that produce multiplexity, and the repetition and segregation of multiplexity, are considered. Social differentials in multiplexity are then discussed. Last, whether multiplexity encourages contact among friends is explored.

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

What Is Multiplexity?
There is no single accepted definition of multiplexity in the social networks literature. Gluckman coined the word, using it to denote "the coexistence of different normative elements in a social relationship." This means the co-occurrence of distinct roles in a dyad, for example, a father and son who are also employer and employee. When only one role exists, the relationship is uniplex; when two or more exist, it is multiplex (cf. Barnes, b). Kapferer offers a second definition of multiplexity: the overlap of different activities or exchanges in a relationship. A third definition is common membership in organizations, peer groups, neighborhood, or kindred (Wheeldon). Wheeldon calls a tie "multistranded" when two people have two or more memberships in common. Whether defined by roles, behaviors, or affiliations, multiplexity refers to multiple bases for interaction in a dyad. The three definitions above are not distinct: for example, kin have well-defined norms to guide their interaction, and certain exchanges typically occur between them. The frequency of multiplex ties for adults is documented in several studies (Bell and Boat; Kapferer; Shulman; Smith et al.; Wellman, a; Wellman et al.).

1288 1 Social Forces 1 vol. 57:4, june 1979


But there is much more speculation than empirical research on the topic. Writers have considered how multiplexity may vary with population size and modernity (Barnes, a; Frankenberg; Srinivas and Bteille), how it may influence community structure (Fischer et al.; Wellman, b), and what instrumental and emotional consequences multiplex ties have for individuals (Boissevain; Craven and Wellman; Mitchell). Multiplexity is a structural feature of dyads (two people who have a social relationship). In turn, dyads are important components of social networks (Barnes, b). Social network analysis examines the structure of relationships among several (or many) people, and the causes and consequences of that structure (cf. Barnes, b; Mitchell). The relationships may be enumerated and analyzed from the perspective of individualsthus, egocentric networksor of a population. The distinction between dyads and networks is important in principle, but in empirical research on social networks, both are often studied. Understanding the structure of small units like dyads is an important precursor to analyzing larger units.

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

Multiplexity in Adult Friendships


Compared to friendships among children and college students, little is known about adult friendships, either as dyadic relationships or as social networks. (Comprehensive reviews of research on children and college students appear in Berscheid and Walster; Lindzey and Byrne.) Recent research in Detroit, Toronto, and a West German city are remedying this, by studying adult friendships for large populations (Laumann; Shulman; Verbrugge, a; Wellman et al.). These studies query respondents about their best friends, the friends' social characteristics, and interaction with them. In the Detroit and West German studies, respondents were asked to name three friends and to indicate if they were also kin, neighbors, or coworkers. This allows us to measure multiplexity of affiliations (friend, kin, neighbor, coworker) in the conditional way described in the Introduction. What motivates adults to develop multiplex friendships? Why do some adults choose their kin, neighbors, and coworkers as close friends? Three factors probably encourage this kind of multiplexity: (1) frequent opportunities for contact, (2) preference for the special similarities that kin (igeighbors, coworkers) have with each other, and (3) the desirability of diffuse and holistic ties. Consider each factor: First, prior studies show that spatial proximity encourages friendship formation (Festinger et al.; Newcomb). Neighbors and coworkers have ready access to each other. This enhances their opportunities for face-to-face contact and development of intimate ties. Second, some interests, memories, and problems are specifically attached to kin, neighbor, and coworker statuses. Relatives share cultural

Adult Friendships 1 1289


traditions and family history. If they have ever lived together, they also share numerous experiences and memories. Neighbors may share an interest in neighborhood affairs. Coworkers share some occupational and "company" concerns. These special similarities may be attractive features, boosting the chances that a kin, neighbor, or coworker becomes a friend compared to acquaintances without these similarities.l Third, multiplexity expands the emotional and instrumental exchanges between two people. Urban researchers have shown that kin, neighbors, coworkers, and friends tend to provide different services to an individual (Adams, a, c; Litwak and Szelenyi). Kin relations are typified by positive concern and interest in each other's activities. Kin are expected to help in major crises and in rites of passage. Neighbors are expected to help each other for small tasks, especially urgent ones. Coworkers are expected to be cooperative and helpful at work. Friends provide empathy, affection, and consensus, and also help for non-urgent needs. From an individual's perspective, each of these affiliations generates obligations, but it also promises needed services and support. When a friend is also a kin (or neighbor, or coworker), two sets of norms are superimposed. People are involved in more of each others' domains than for a uniplex tie. Offering more instrumental services and emotional supports to each member, the multiplex tie is more diffuse and holistic, and it feels more secure to the two members. We shall call this aspect of multiplex ties "role accretion." These three factors can be placed in a more general theoretical perspective about friendship formation. Which people become close friends in a population is influenced by contact opportunities and by preferences. Pairs of adults have different chances of ever meeting and, even if they meet, different chances of nourishing their acquaintance by face-to-face contact. In addition, pairs of acquaintances are more (or less) attracted to each other and make greater (or lesser) efforts to get together. In our discussion of multiplexity above, the first factor refers to contact opportunities; the second and third to preferences that people use, so that only some acquaintances become close friends. The sociologiet's task is to describe nonrandom pairing and then to determine the factors which cause it. 2 Unfortunately, the data used here do not allow measurement of each factor separately, to gauge their relative importance in the formation of multiplex ties. We have the structural outcome of the processes, not the processes themselves. Interpretation of results must therefore take all three factors into account.

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

Data Source
Two cross-sectional sample surveys of adult populations are used: the 1966

1290 1 Social Forces 1 vol. 57:4, june 1979

Detroit Area Study and a 1971 survey in a West German city (given the pseudonym Altneustadt). The Detroit study is a multistage, probability sample of 1,013 white males, ages 21-64. Respondents had to be born in the United States or Canada, be members of a primary family in the sampled dwelling unit, and be residents of the Detroit Area Study's sampling area, which is roughly the Detroit SMSA. The Altneustadt study is a single-stage, systematic probability sample of eligible voters in 1971. The voter register contains all German citizens 18 years or older who are municipal residents. The sample has 820 cases and included both men and women. Further sampling details are given in Schuman and Verbrugge (a, 53-6). In both surveys, a respondent was asked to name his or her three best friends, apart from household members. In the Detroit sample, only nominations of male friends were allowed. In Altneustadt, respondents could name male or female friends. For each friend named, the respondent was asked if the person was a kin, neighbor, or coworker. 3 In this paper, the entire Altneustadt sample is used, not just males. Analyses were performed separately for males and females. Few differences by sex appear (Verbrugge, a, Chap. 6), so the sexes have been pooled. Several limitations of the data sets must be mentioned: First, the friendship dyads are generated by one member of the pair, the survey respondent. Thus, the dyads center on these people (egos). In real life, friendships are mutually formed and maintained. 4 The perspective here is not entirely faithful to reality, but it is necessary since the sampled units are individuals, not friendship dyads. Second, the question eliciting names of friends does not ask ego to name the best friend first, the second-best friend next, etc. However, Verbrugge (a, Chap. 2) shows that most egos do name friends in order of their affection for them. Thus, we consider Friend 1 (the first-named friend) as best friend, Friend 2 as next-best friend, etc. Third, the eliciting question for friends asks for intimates who are seen frequently, not intimates regardless of contact frequency. How respondents considered both factors (intimacy and contact) when naming their friends is unknown. It is an unfortunate restriction on the sample of friendship dyads, but there is no way to remedy it. Fourth, the Detroit data do not have full information on multiplexity. In their nominations of friends, some respondents named brothers (4% for Friend 1, for Friend 2, and for Friend 3) and brothers-in-law (6% for each Friend). These respondents were asked to replace such nominations by names of other people (not brothers or brothers-in-law). Thus, there are two sets of friendship data for Detroit: first-mention and second-mention. For the brother and brother-inlaw friends named, coworker status is known, but not neighbor status. Thus, the first-mention data cannot be used to study neighbor multiplexity. Both data sets were analyzed for the paper; the text and tables report results for the most appropriate set. Fifth, the definition of neighbor status

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

Adult Friendships 1 1291


differs in the two surveys (see note 3). In Detroit, friends living within a 10minute travel distance are coded as neighbors. In Altneustadt, friends who live within the municipality are coded as neighbors. The municipality is small and most of these friends do live within 10-15 minutes of the respondents, but it must be acknowledged that some do not. Sixth, readers should be aware that "coworker" is defined quite generally. Two people who work for the same company but have very different jobs may be coded as coworkers. The two data sets are used for replication. If the structure of multiplex friendships, their social correlates, and their consequences are similar in two sites, we have some confidence in the generality of our results. We are more sure that prominent and pervasive features of adult friendships have been foundat least for Western industrial settings. In fact, the analysis does show similar results for the two sites. All sizable differences are noted in the text, but such differences are rare.

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

Measuring Bias in Friendship Choices


Randomness in social relationships means the absence of social structure. Nonrandomness is a necessary feature of structure. To say anything important about multiplexity in adult friendships, we need to demonstrate at least that it appears nonrandomly. In the analysis, we show bias (nonrandomness) in two aspects, called repetition and segregation. We first prepared cross-tabulations which show the affiliation of Friend n (n = 1,2,3) on one stub and of Friend m (m n) on the other. For each affiliation, three tables were produced: Friend 1 with Friend 2, Friend 1 with Friend 3, Friend 2 with Friend 3. For example, neighbor status of Friend 1 was cross-tabulated with neighbor status of Friend 2. This resulted in nine basic tables for each site. To demonstrate a bias toward (or against) repetition and segregation, we must compare frequencies of observed dyads with expected frequencies. For this paper, the expected frequencies are based on a random-choice model, which assumes that people choose friends at random from a pool of eligible friends. The larger the difference between observed and expected frequencies, the more people are using their opportunities and preferences to select close friends. Two measures of bias are used: an odds ratio and a marginal ratio. The odds ratios computed for the tables compare the odds that respondents with a Friend type i choose another Friend type j to the odds that all other respondents (whose Friend is not type i) choose Friend type j. A brief explanation will clarify this: For a 2 x 2 table, the odds ratio is (F1,F221F12F21) (Mosteller). A table with more categories on either side can be collapsed to 2 X 2 tables in numerous ways to test different

1292 / Social Forces / vol. 57:4, june 1979

hypotheses. For example, neighbor affiliation has three categories (see note 3). To examine if respondents whose Friend 1 is a neighbor also

tend to choose Friend 2 who are neighbors, this odds ratio is computed: F2z (F., F, F+F)l (F^. F^)(F,, F), where i refers to the category "Friend who is neighbor." To test the hypothesis about repeated choice for an I x I table, a total of 1 such odds ratios are computed. For the neighbor table, this means three odds ratios are computed. What are the expected odds ratios for a random-choice model? In all cases, the expected ratio is 1.00, which signifies independence of the two marginal distributions. Observed odds ratios are tested for statistical significance (difference from 1.00) by the procedure detailed in Goodman
(:4-13).

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

The marginal ratio states how likely respondents with a Friend type i are to choose another Friend type j, compared to the number of type j eligibles. The ratio is F 2j /F0 , where F Z, is the observed cell frequency and Fi; is the expected frequency. As the ratio's name implies, expected frequencies are derived from a table margin. What does that choice imply? Ideally, we would like to know the characteristics of people eligible for friendship for the population we have sampled. Then, if a certain group of respondents choose their friends randomly from this pool, their Friends (1,2,3) will mimic the social characteristics of the pool. But there are no data on people eligible for friendship. The best solution is to use the characteristics of the Friends named, and to assume that this closely reflects the characteristics of eligible people. 5 Thus, the expected frequency (Fz; ) is derived from the bottom marginal of the table (F = [F. IN]F1 .). No test of statistical significance for that ratio is available. Overall, the odds ratio has more desirable statistical properties than the marginal ratio. One has been mentioned: the existence of a test for significance. In addition, the odds ratio is invariant with row or column multiplications (i.e. is not sensitive to its margins). This is not true for the marginal ratio (Blau and Duncan; Tyree). This feature of the marginal ratio hinders numerical comparison of tables with different marginal frequencies and even of cells within a table. Tables 1 and 4 for this paper show odds ratios, computed as described above. Marginal ratios are discussed in the text but are not shown in Tables 1 and 4. They are available from the author on request. Tables 2 and 3 also show bias in friendship choices by cornparing observed and expected distributions, the expectations reflecting random choice. The random-choice model assumes that the presence of kin, neighbor, and coworker statuses appear independently of each other in a friendship dyad and in friendship networks, and (for Table 3) in the pool of eligible friends. As before, the marginal distributions are used to derive the expected frequencies. Table footnotes describe how the expectations are computed.

Adult Friendships 1 1293 Repetition and Segregation of Multiplexity


In both Detroit and Altneustadt, kin, neighbors, and coworker are frequently chosen for friends by adults. Descriptive results are: in Detroit, 15-18 percent of close friends are also kin, 41-45 percent are neighbors, and 27-33 percent are coworkers. 6 (The range is based on three percents, one for each of the nominated Friends.) In Altneustadt, 21-23 percent are kin, 55-66 percent are neighbors, 7 and 18-26 percent are coworkers. Multiplexity is so common that 63-68 percent of the Detroit friendships are multiplex in one or more of the three affiliations, and 75-80 percent in Altneustadt. In the network of three friends, only 12 percent of Detroit respondents name no kin, neighbor, or coworker at all, and a mere 6 percent of Altneustadt respondents.
REPETITION OF KIN, NEIGHBOR, AND COWORKER ACROSS FRIENDSHIPS

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

People who select a kin, neighbor, or coworker as one close friend might repeat that criterion for other close friends. In other words, the opportunities and preferences which urge multiplex ties probably influence a person's choice of all close friends. The hypothesis is tested in two ways: First, take the three pairs of friends (ego with Friend 1, with Friend 2, with Friend 3). Does the multiplexity status of one pair tend to be repeated for the other pairs? Table 1 shows odds ratios appropriate for the question. Second, examine the friendship network (all three friendships as a unit). Are several of the respondent's friendships multiplex? In Table 2, the observed frequency of 0, 1, 2, and 3 friends who are kin (neighbor, coworker) is compared to an expected frequency, based on a random-choice model. In Table 1, diagonal cells indicate repetition of a criterion in two friendships. All of them have odds ratios and marginal ratios greater than 1.00, indicating a tendency to repeat multiplexity. All but one of the odds ratios are significant (P < .05). Note that even the specific type of kin tends to be repeated across friendships: if a same-generation kin is chosen as Friend 1, there is a strong possibility that Friends 2 and 3 will be same generation kin. Similar results appear for the residential distance between respondents and friends. Most (85%) of the off-diagonal cells have odds and marginal ratios below 1.00. 8 These off-diagonal results are not simply redundant with the diagonal ones: (1) They reveal that when ego chooses one type of kin for a friend, he or she is unlikely to choose another type of kin for the other friend. In other words, an individual's preference for kin friends is specific to one type of kin. (2) Most odds and marginal ratios above 1.00 are for friendships with a same-generation kin and an upper-generation kin. This suggests some compatibility in having friends within on s family of

1294 1 Social Forces 1 vol. 57:4, june 1979


Table 1. REPETITION OF MULTIPLEXITY IN PAIRS OF FRIENDSHIP DYADS, DETROIT, 1966 AND ALTNEUSTADT, 1971
(Odds Ratios)* Detroit Kin Friends t Not Kin Upper Not Kin FD1 x FD2 FD1 x FD3 FD2 x FD3 Upper genen Lower Altneustadt Not Kin Upper Same Lower

Same

5.84e 2.60# 4.75# .21 .94 .54

.24 .31 .29 10.23{ 4.19 14.62# 1.76 2.61 1.86


-

.19 .57 .29 3.49 .37


-

.25 .14 .09 1.74


-

9.73* 4.72# 8.194 .15 .16 .11 .21 .30 .18 .o6 .35 .30 Neighbor

.19 .94 .37 27.15# 3.40 9.20# 1.46 .82 1.47


-

.19 .22 .16 1.91 5.58 4.73 6.50 4.24* 6.34# .80 .58 .58 Farther

.07 .18 .10 .73 .89 1.36 155.11# 36.90k 48.67#

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

Same gener. .22 36 .22 Lower gener. .29 .29 .21 Neighbor Friends ll Neighbor Neighbor FD1 x FD2 FD1 x FD3 FD2 x FD3 Local area .58 2.28 .52 Farther 39.31# 14.29# 77.58# 5.08# 2.20# 4.76# 1.64 4.55 2.58

Area

Area

3.381 2.544 3.544 38 54 39

.37 .55 .42 3.75# 2.28# 3.18# .48 .60 .51 Not Cow

.65 .49 .41 .40 .58 .57 7.20 6.76# 7.34#

6.53# 4.254 4.23* .34 .41 .45 .14 .23 .26 Coworker

.34 .57 .55 4.60# 2.85# 3.71# .80 .63 .47 Not Cow

.22 .25 .25 .71 1.02 .63 11.36* 7.55* 8.83#

Farther 50 .43 50 Coworker Friends ll Coworker Coworker FD1 x FD2 FD1 x FD3 FD2 x FD3 Not coworker 4

4.61# 2.69# 4.104 .22 37 .24

.22 37 .24 4.61 2.69 4.10

5.17# 3.82# 5.65# .19 .26 .18 (C-1)

.19 .26 .18 5.17 3.82 5.65 in a table are

*All ratios are computed and shown, although only (R-1) independent of each other.

tFirst-mention data for Detroit (includes brother and brother- in-law) For secondmention data (which excludes them), the ratios are similar. #Significant at P < .05 Only diagonal cells are tested for statistical significance. Critical values are in Goodman (Table 3, right column). "Upper gener." means one or more generations above the respondent. "Same gener." means same generation. "Lower gever." means one or more generations below the respondent. I"Second-mention data (excludes brother and brother-in -law). #Tables have one degree of freedom, no only one odds ratio is tested for statistical significance. -Ratios are not computed when a cel) has zero cases.

Adult Friendships 1 1295

Table 2. REPETITION OF MULTIPLEXITY IN THE FRIENDSHIP NETWORK, DETROIT, 1966 AND ALTNEUSTADT, 1971 (IN PERCENT)
Detroit* Observed Number of ego's friends who are kin 0 2 3 N SIG?t Probabilities to generate expectations: Friend 1 Friend 2 Friend 3 Number of ego's friends who are neighbors 0 1 2 3 N SIG? Probabilities to generate expectatioos: Friend 1 Friend 2 Friend 3 Number of ego's friends who are coworkers 0 1 2 3 N SIG Probabilities to generate expectations: Friend 1 Friend 2 Friend 3 Expected+ Altneustadt Observed Expected

64.4 58.0 34.6 25.5 7.3 6.9 2.8 .5 100.0 100.0 1,005 SIG (X2 = 137.6)

44.1 58.0 20.8 41.5 13.0 12.8 1.4 8.4 100.0 100.0 760 SIG (X 2 = 367.2)

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

1757 1499 1710

2513 2342 2303

30.8 32.5 20.7

19.1 42.2 31.1

16.0

7.6

100.0 100.0 1,005 SIG (X2 = 225.5)

7.2 18.4 21.6 30.9 42.8. 27.8 32.2 19.1 100.0 100.0 760 SIG (X 2 = 261.8)

4518 4073 4133

.6579 .5803 .5000

46.2 29.2 15.4

9.2

100.0 1,005 2 = 267.2) SIG (X

35.5 43.9 18.1 2.5 100.0

51.4 62.4 21.7 38.4 11.6 9.4 4.3 .8 100.0 100.0 414 SIG (X2 = 116.5)

.2687 .2975 .3085

2367 1881 1714

*First-mention data are used for top panel (kin), and second-mention data for other panels (neighbor, coworker). tTo get expected percents, multiply probabilities (p or 1-p) shown in a panel, for each permutation of the three friends. There are 8 of these altogether. For example: kkk (no kin); kkk, kkk, kkk (one kin); etc. Probabilities are added to obtain the expected probability for "one friend who is . . . and for "two friends who are. . . .. The probabilities are converted to percents in the table. (P tSIG means the observed and expected distributions differ significantly .05, df = 1-1, where 1 = number of categories). Currently employed respondents only.

1296 1 Social Forces 1 vol. 57:4, june 1979


orientation, such as a mother and brother. (3) Having kin friends just one generation apart is more likely than having kin friends two generations apart. Two other results are similar to (1) and (3) but refer to neighbor friends. (4) Proximity preferences are very specific. For example, if ego chooses a "local area" resident for friend, he or she is unlikely to choose someone living any other distance for the other friend. All but two of the odds and marginal ratios for the off-diagonal cells are below 1.00. 9 (5) When a person has a neighbor friend and a nonneighbor friend, the latter is more likely to live in the local area rather than far away. Table 2 shows there is a bias toward choosing no friends who are kin (neighbor, coworker) or all three friends who are. Stated differently, people tend to use these affiliations persistently when choosing friends, or not at all. As a result, having only one friend who is kin (neighbor, coworker) occurs less often than a random model predicts. Having two such friends usually occurs less than expected also. In summary, both data sets show a pervasive tendency to accumulate

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

multiplex friendships. Having one such friend, an adult is likely to have others.
The converse is: When a friendship is uniplex, the others are probably uniplex too. The intriguing question which these data cannot answer is whether repetition is due to contact opportunities or preferences. (1) People live in limited social arenas which influence who they become acquainted with and who they see routinely in their daily rounds. These limitations should influence similarly how ego develops all his or her close friendships. Moreover, a chain of friendship formation may occur. For example: If ego has a neighbor friend, and that person likes another neighbor greatly, ego may also develop a close tie with that third person. (2) People vary in their preference for the special similarities and services that kin, neighbors, and coworkers offer, and in their preference for holistic ties. These preferences probably influence the development of all close friendships too.
SEGREGATION OF KIN, NEIGHBOR, AND COWORKER IN FRIENDSHIPS

So far, we have considered the addition of just one affiliation to a friendship. Friends may be more multiplex, overlapping two or all three affiliations (for example, a close friend who is also a coworker and cousin). These highly multiplex ties may be very attractive since they promise diffuse rewards. On the other hand, highly multiplex ties are likely to be rare in industrial societies, which are characterized by role segregation and high residential mobility compared to non-industrial societies. Kin are unlikeiy to be neighbors or coworkers; coworkers are unlikely to be neighbors or kin; etc. Thus, the pool of eligible friends with multiple affil-

Adult Friendships 1 1297

iations is very small, and the chances of finding such friends, even if they are desirable, is also small. Table 3 compares the observed percent of friends who are multiplex in one, two, or three affiliations with the expected percent from a randomchoice model. Consider the highly multiplex ties (with two or three affiliations): of 24 entries in the table, 75 percent show that observed ties are less than expected. There is a bias against having friends who are highly multiplex. On the other hand, the observed friendships with one affiliation exceed expectations. One, and only one, criterion tends to be used in choosing friends. Does this mean that adults are actually reluctant to have highly multiplex ties, because they are too demanding and "sticky"? (If so, role accretion is not a reason for developing multiplex ties, but a reason to avoid them!) Or does it reflect very limited opportunities for having acquaintances with multiple affiliations? Although the survey data do not allow us to answer these questions directly, some evidence against the first explanation is in Table 4. That table examines repetition of multiple criteria across friends. All ratios for diagonal cells exceed 1.00, and all of the diagonal odds ratios are statistically significant. The specific kind of multiplexity in one friendship tends to be repeated in others, even when multiple affiliations are involved. For example, people who choose a kin-neighbor-coworker friend are 3 to 19 times more likely to choose another such friend than expected. A majority (69%) of the off-diagonal cells have odds and marginal ratios below 1.00. This indicates a bias against using different sets of criteria for friends. (The exceptions tend to be patterned. They are discussed in Verbrugge, a, 473-77.) It is appealing to interpret the specific repetition shown in Table 4 as a product of preferences: people with a highly multiplex friendship enjoy it so much they try to duplicate it in other friendships, despite the problem of a small pool of eligibles. Role accretion appears desirable. This interpretation is cautionary, since opportunities can produce this specific repetition for some groups whose acquaintance pools are unusually multiplex. For example, recent immigrant groups tend to have highly multiplex ties; these immigrants are often kin, live in a particular neighborhood, and work in the same companies (Breton). Such pools are not common in industrial societies. In summary, although preferences are not the sole explanation for repeated high multiplexity, they are certainly an important explanation and probably the principal one. If preferences encourage highly multiplex ties, then segregation of affiliations (Table 3) is due to limited opportunities for having such friends. In fact, if we had a correct expectations model showing the real-world probabilities of the categories in Table 3, the expectations for two and three

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

AND ALTNEUSTADT, 1971 (IN PERCENT)

VN
41

N
Q

Number and Combination of Affiliations in a Friendship


u
N

Ow
W

Z ^

o Z
C LL
M

C^ Q

w a Z Z
Friend 3
W
C +

Table 3. COMBINING KIN, NEIGHBOR, AND COWORKER AFFILIATIONS IN CLOSE FRIENDSHIPS, DETROIT, 1966
o
(O (O i
U

m U

Z F Z_ O
Altneustadt
V X M 41 O

LL
W

1
Friend I Friend 2
O L

f,/'j

4 c Y

C
0 .O O

Detroit*

AItneustadt Detroit
O W

Altneustadt
X W

Detroi t Obs. Obs.


p

o n

0L

u n m X lJ IG

0 -

41 a

Obs.

Exp.t

Obs.
O

Exp.

Obs.

Exp.

Obs.
N

Exp.

Exp.

Exp.

N '

None
1

...

tV

M ^ M
P. M OC

s^

ff

25.3 3.2 40.0 15.5 5.2 1.0 9.7


MNO LnN 01^O

SOIQ\NLACVA9OO

M M Ol Y^? r .- O

0 I I I 2 2 2 3
;

Coworker Neighbor Kin Nbr, Cow Kin, Cow Kin, Nbr Kin, Nbr, Cow
e('-1- aD ^ 001

0LA MN MQI^N 0

N Ipso en N ^ ? Q1 0 Op

en,D O N ^ ^ IO O ene.4 ^ O

1^ ^u1NO4G^IO t^

NMN 1^^0^ N O M

UA? LA (.4-t CD en M O N jC

I-.I-.I`.00 u1O 0

M ('4 (0 .- S 0(0 LA O N TC

O LA ^ ^ a0 LA 9O S 0 .41: 4j,4:, Ql O 1^ U

O ^ I^MO N^^N en.en--

OMI^OM^^ O O M. O ?

U^MO U\u\^ O c .- o fV ?^ OQ

34.5 37.2 16.4 13.7 30.6 32.5 3.2 2.9 11.2 9.4 .8 1.1 2.4 2.9 _._9 .3 100.0 100.0 1,005 2 51G (X = 17.8)
O M ^ 1^ N

S^pN M01 N IOVA V

1^ 01C0 N 01 ^ IO 1^"

S N N M O'0 I^

N 51G7* Probabilities to
('-N
O

3 N uo u 0 ^ 0.oO LA N^

generate
A? 90N. N 0(0 a0

.teno J

expectations:

3 3 L L

< o
.I

100.0 691
N_ )

30.4 3.2 37.2 10.3 3.9 1.1 12.6 1.3 100.0


II

eO N

18.9 22.3 3.7 3.3 44.1 42.9 11. I 7.5 8.2 6.4 1.1 .5 14.4 13.0 2.1 .5 100.0 100.0 760 2 51G (X 33. I)

^ O M

00

37.3 39.4 19.2 16.9 28.8 26.9 2.1 2.2 11.5 9.4 1.0 .9 1.8 1.5 .6 _._5 100.0 100.0 997 N5 (X 2 = 10.0)

22.0 27.3 3.1 3.5 44.1 40.7 14.1 8.7 5.8 4.7 1.0 5 9.6 13.0 _._4 1.5 100.0 100.0 737 51G (X2 50.8) 32.4 36.8 22.1 17.8 29.1 25.7 3.5 2.9 9.2 12.4 1.0 1.4 2.6 2.0 __ ._1 1.0 100.0 100.0 954 51G (X2 = 38.5)

5 I G ( X2 = 40. 7l

'

011 ^ ^ N ,A O V\ O , N 'O ^

1298 1 Social Forces 1 vol. 57:4, june 1979

Neighbor Coworker
wao

*Second-mention data must be used, since coworker status is not available for first-mention data.
41

t5ee note b, Table 2.


v 4) L
(0
4.

00._444.L L L o o Z C ^ Y L 00 2 Z u (0 0 0

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

* .- L O 10 L Y l^! L W LA. C 3 C L C C C .41. M0 L 4 09 0 ) Ol O Q O d - d - - - h.O C O. C .O Z u Z Y Z Y Y Y u O N X- d z o O^^ ^NNNM Z VI 1 OOY U Z O


N1^O M en LA

Kin
N LA ^

^

LAoen

Ol 11 MYAOI

en enen

.0716 .4507 .2687


O? N J L 4-

.2513 .6579 .1289 .0532 .4052 .2999


O ? N 0 C

.2415 .5984 .1031

.0723 .4109 .3249


c E E
0

.2533 .5499 .0955


Melen ('40-t I'--N N O? M V

41

a+

t,

> L N L 0

4+

(0

0 u

0 Y-

014

41 u

U C 6

L N

v 4-4.041 u

U >. LA 4a+ 0 V-

VI L L L .0 y L

.0 0

- N N in t 41 41 C '-4

mOvu
3 40 c

c E O O

04 C . O

c m G 0

N C

ai .-< -

W V T > N a+

a1 41

41 a (u u >

04- 4. L y 0.0.0

E c04)0 L 0-

a+

44

E 41

01.'

0a W

1)

N G)

04.4(9-

E a L N .n a ai O ^ ^ o- L . wc H m

d 49 W O L ^

m y .Ot

4) L L w 0

00

0 ^``w Y) -N-4. 4. u0wm

.0t- oL

c N - v O

V 00.C N 4

X O I N 4O

O 00

m c u v

*51G means the observed and expected distributions differ significantly (P::': .05). N5 means not significant. The probabilities are the proportion of Friend i (i = I, 2, 3) who have the affiliation named. Kin probabilities for Detroit differ from those in Table 2 because second-mention data are used here. Coworker probabilities for Altneustadt differ from those in Table 2 because all respondents are used here (not just currently employed ones). Other values may differ slightly from Table 2 due to different handling of N.A. cases.
p L C C L
w II C N Z

c rn
W---t< w N W u C en.,., y OI (000

M 404 Q) L L 000 9-4L 4 Y 0 V V


L Ot N L LA. 041.0

a+ o u 4n c 4)00

v - E w I 00 N t* O - C

N 04-0>.

Adult Friendships / 1299


affiliations would be smaller. Differences between observed and expected frequencies would reflect preference factors. We suspect the ratios would then exceed 1.00, showing the preference for high multiplexity. Whatever the reasons for segregation of kin, neighbor, and coworker statuses, it is demonstrated consistently in the data. Friendships

tend to be multiplex in only one of the affiliations, and not more.

Who Chooses Kin, Neighbors, and Coworkers as Friends?


Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

Some social groups tend to choose kin, neighbors, and coworkers as friends more than other groups do. This reflects their different opportunities for contact with eligible friends, and their different desires and needs. Which social groups have restricted social contacts, preferences for special similarities offered by kin, neighbors, and coworkers, or desires for holistic ties? We expect kin friends to be frequent in social groups which need frequent economic assistance and which value communal, expressive behavior. Examples of the former are people of low socioeconomic status and the elderly. Neighbor friends should be common among people who need to minimize costs (time, money, physical movement) to see friends or who have limited social contacts outside the neighborhood. For example, Rosow finds that elderly adults are locally focused in their friendships. This may be due to transportation costs and limited social contacts, since they have no current coworkers and often few kin living nearby or of similar age. Coworker friends are probably preferred by adults deeply involved in their work. Coworkers can discuss work details and empathize about work problems better than other acquaintances. People who are very involved in their work may have limited contacts elsewhere and become quite dependent on coworkers for friendship. The timing of work hours may also cause some occupation groups to recruit friends from their worksites more than elsewhere; for example, workers on night shifts. Opportunities, special similarities, and role accretion do not have equal importance in explaining the choice of kin, neighbor, and coworker friends. The main motivations for choosing kin as friends are probably the extra roles they bring to the tie and also special background similarities. Opportunities (easy access and limited outside contacts) are probably most important in choosing neighbors as friends. Similar interests and limited contacts away from work motivate choice of coworkers. Differentials in choice of kin, neighbor, and coworker friends were examined for these respondent characteristics: occupation, employment status (self-employed or not self-employed) occupational prestige (Treiman), education, nationality (Detroit only), political preference, religious

Table 4. COMBINING KIN, NEIGHBOR, AND COWORKER AFFILIATIONS IN CLOSEFRIENDSHIPS, DETROIT, 1966 AND ALTNEUSTADT, 1971 ('IN PERCENT)

t.IJ 0 0
t/)

....

(Odds Ratios)* Oetroit t Altneustadt knc knc knc knc knc knc knc knc knc knc knc knc
knc

0 n

knc

knc

knc

IlJ

....
.76 .83 .63

III

4.22 .96 .79

0 '1 n

'T.l

knc FOI x F02 FOI x F03 F02 x F03 knc .73 .80 .49 1.87 2.36 .41 .40 .41 10.64* 9.68* 8.61 * 2.91 3.66 2.07 .38 .52 .38 1.04 .89 1.06

* 5.03* 3.29* 4.84

1. 27 1. 16 1.67

{Il

< 0
2.57 .68 2.11 3.84 .31 .68 .53 1. 43 * 3.62* 1.96 4.00* 1. 31 2.28 .75 .51 .85
I. 19

knc .71 .63 .69 1.29 1.29 .48 .78 .49 1.02 .76 .52 .60 .42 .47 .24 .26 .56 7. 21. 5.22* 5.93 * 9.94* 7.96 37.68 l .82 .18 .29

1. 03 .93 .94

* 4.74* 2.93* 3.84

VI 'I

1.66

.42 1.03 .73

1.25

knc

~ _.

III ==

=
....

knc

.80 1. 17 1.24

. 56 .52 .37

.41 .67 .49 1.30 1.09

.28 .49 .33

.38 .53 .54

.31 .37 .28 .66


I. 20

'-0 'I '-0

knc
I. 29 3.03 .47

I. 10

1.13 12~45* .54 .26 .82 3.19: 3.55* 3.15 1.34 .74 .67 1.65 2.16 7.97l
I. 55 I. 97

.73 1.01 .53 .57 .58 .67 .76

.90 .24 .82

l 19. 34* 7. 14* 9.26


.66 .69 .37 .36 .38 .30 .46 .59

.34 .45 .38 6. 38 3. 83* 4.55

1. 24

knc .75

.39 .40 .25

.57 .82 .64

1
.98 .76 .17 .27 .30 .86 1. 25 1.40 1.00 I. 10 .55

.87 1. 02 .90

knc

.98 1. 12

.85 .83

10~1O*

10.7 4: 4.06* 7.15

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

*The stubs use the letters k, n, and c to designate kin, neighbor, and coworker statuses, respectively.
0
.4)

A bar over a
a v
^ U 0 N

F-

0.

Y 04

vv

04-

0.':

C N 1

'-'(0 m ++

O 3 ^ O U u

0--

L C

u+m

04.0

00)

<40

a c

L ._

v 0 L > 0 > O 0

letter means "not."


N L
a O1 l ^N

-.0 O

.00 0(0

0 0jA L. d .4)0

7 V- V 4-0 u

>.0 0) > V > G --0 Y a 00

.- x

m uwcm

v a c v w 0(0 c u

V L O E rn m E 0 m

d 0 0 m - o

(DC W N oj o t ja

04.0 O C rn . '4-

-0 m v

rw-t(
N 0 0 N
C

2 VI t v a c

(Table 3. right column).


E ja
v

o c c 0 44 a L

F L Y- wF W E 0

n C c m 0 c L

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

m
j0 -
C

All ratios are computed and shown, although only (R-l) (C-l) in a table are independent of each other.

tSecond-mention data (excludes brother and brother-in-law).


Only diagonal cells are tested for statistical significance.
Y O T
C

o u a v

*Significant at P S .05. Critical values are in Goodman


U (p >

Cl
N
U N J

u U

04

rn

u_

U)

u c

C 004

v
w
;

-Ratios are not computed when the cell size for two friendships (e.g., knc with knc) is 2 or less.
u
N W v O

ja

04 0. E u

w m

'u iC

01

> l:l.o

... Adult Friendships 1 1301

r::

1302 1 Social Forces / vol. 57:4, june 1979

preference, residence length (Altneustadt only), age, sex (Altneustadt only), and marital status. The results confirm the hypotheses above (data not shown). These differentials are most pronounced: Compared to the sample average, kin friends are especially selected by production workers (craftsmen, operatives, laborers), people of low prestige, and the self-employed. As age increases, so does the choice of kin for friends. Catholics, South Europeans, and previously married women are most likely to have kin friends, compared to other religious, ethnic, and sex-marital groups. Explanations in terms of economic and emotional needs and valuation of family and kin ties can be stated for these differentials. Neighbor friends are most common among production workers, compared to other occupation groups. Education and occupational prestige are inversely related to the percent of neighbor friends. Increasing education prestige, and professionalism probably give individuals more social contacts and more diverse ones. Friendship choices are less limited to one's local area than for persons of low socioeconomic status. In both sites, elderly people are more likely to choose neighbor friends than other age groups, reflecting their narrowing social arena (Cumming and Henry; Rosow). Professionals, administrator/managers, service workers, and production workers tend to select coworker friends in Altneustadt, more than other occupation groups do. So do workers of high and low prestige in Altneustadt, compared to middle-prestige groups. These results may reflect high work involvement for upper status groups and limited social arenas due to timing of work for lower status groups. In Detroit, there are no consistent differentials in the occupation or prestige groups who choose coworker friends. In both sites, coworker friends are especially uncommon for the oldest and youngest adults. This may be due to opportunities for the elderly (most are retired) and to preferences of young adults (many still have childhood friends and may prefer them to coworkers). Coworker friends are most common at ages 40-49 and 50-59. Married people have coworker friends more than single and previously married people do. The finding is suggestive: if seeking a mate, nonmarried people nourish diverse acquaintances, in and beyond their worksite. Married people may be more content to find friends from accessible coworkers. In summary, choices of kin, neighbors, and coworkers as close friends fit the hypothesized patterns. Detailed discussion of these social differentials in friendship choice appears in Verbrugge (a, 478-87).

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

Adult Friendships 1 1303 A Consequence of Multiplexity: Contact Frequency


Writers have speculated that multiplex ties are strong but also "sticky" (Frankenberg; Kapferer; Wheeldon). The more multiplex a tie, the more readily its members are mobilized for support, but the less easily they can "withdraw completely from contact with one another" (Mitchell, a, 24). What empirical consequences does multiplexity have for behavior between friends? Does it increase how much two people like each other, how many goods and services they exchange, how often they get together? The Detroit and Altneustadt surveys have data on contact frequency of friends (away from work) but not on other interactions. 10 How should kin, neighbor, or coworker status influence contact frequency of friends 11 Consider the effects of opportunities, social similarities, and role accretion. First, opportunities for contact differ sharply for the three multiplex ties. Literature on contact frequency among friends and relatives demonstrates that distance between them is a strong predictor (Adams, b; Klatzky; Litwak; Reiss; Roethlisberger and Dickson). Having a neighbor as friend is likely to increase contact sharply. Coworker friends may also see each other more than average, since work contact gives them ample opportunity to plan off-work activities. Kin friends have no routine access to each other by virtue of kinship alone and may be very distant geographically. If any opportunity effect exists for kin, it is probably negative, so that kin friends see each other less than average. Second, the attraction of similar background and interests should encourage contact in all three multiplex ties, compared to contact with people who are "fust friends." This should equally affect kin, neighbor, and coworker ties. Third, role accretion should boost contact more for neighbors than for kin (whose obligations are infrequent) or coworkers (whose special services are given at the worksite). Together, the net effect of these three factors should be to increase contact frequency for kin, neighbor, and coworker ties. The effect should be strongest for neighbor friends, and weakest for kin friends. In the two surveys, contact frequency is coded as days per year that the respondent and friend see each other away from work. Contact frequency (Y) was regressed on a set of dummy variables for kin, neighbor, and coworker status (Table 5). The hypotheses for neighbor friends are supported: Neighbor friends see each other much more than non-neighbor friends. In Detroit, their contact frequency is 16-20 days per year above average; in Altneustadt, 14-18 days. (Separate regressions are estimated for each Friend 1, 2, and 3; the range of days refers to three regressions.) Coworker status has a small positive effect in Altneustadt, but no effect in Detroit. Kin status has no persistent effect in the two sites. The coefficient for kin is sometimes positive, sometimes negative, and always small.
?

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

1304 1 Social Forces / vol. 57:4, june 1979


Table 5. EFFECTS OF KIN, NEIGHBOR, AND COWORKER STATUS ON CONTACT FREQUENCY OF CLOSE FRIENDS, DETROIT, 1966 AND ALTNEUSTADT, 1971*
(Days per Year) Kin Detroit FD1 FD2 FD3 Altneustadt FD1 FD2 FD3 Neighbor Detroit FD1 FD2 FD3 Altneustadt FD1 FD2 FD3 Coworker Detroit FD1 FD2 FD3 Altneustadt FDI FD2 FD3 Average Kin Not Kin N R2 SIG?t

60.5 54.8 54.1 74.5 63.0 59.1

+12.2 +2.6 -2.9 +0.1 -3.7 -2.4

-1.0 -0.1 +0.2 -0.0 +1.2 +0.8 Area Farther

996 992 945 759 745 689 N

.003 .000 .000 .000 .001 .001 R2

NS NS NS NS NS NS 5107

Average Nbr

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

60.5 54.8 54.1 74.5 63.0 59.1

+20.1 +16.4 +16.5 +13.8 +15.6 +18.0

-12.7 -8.6 -7.6 -7.5 -7.8 -8.6 Not Cow

-29.5 -20.5 -26.6 -50.8 -40.0 -36.5

996 992 945 759 735 689 N

.100 .060 .065 .142 .136 .143 R2

SIG 510 SIG 510 SIG SIG 510?

Aver ,e Cow

60.5 54.8 54.1 70.5 61.6 57.4

-2.6 +1.9 +1.9 +14.3 +8.1 +18.4

-1.0 -0.8 -0.9 -4.4 -1.9 -3.8 Not Kin Nbr Area Farther Cow Not Cow

996 992 945 413 403 384 N

.001 .000 .001 .018 .005 .021 R2

NS NS NS SIG NS SIG SIG?

Kin,Nbr,Cow Average Kin Detroit FD1 FD2 FD3 Altneustadtt FDI FD2 FD3

60.5 54.8 54.1 70.5 61.6 57.4

+13.0 +2.6 -0.5 +2.9 +7.2 +11.9

-1.0 -0.1 +0.0 -0.8 -2.0 -3.5

+20.1 +16.6 +17.2 +11.6 +14.4 +18.0

-12.7 -8.9 -8.3 -4.2 -2.3 -5.9

-29.7 -20.3 -26.2 -37.4 -38.6 -34.4

+0.1 +3.4 +4.7 +10.4 +3.8 +17.2

-0.0 -1.5 -2.3 -3.2 -0.9 -3.6

996 992 945 413 403 384

.106 .061 .068 .101 .119 .152

SIG 510 SIG 510 SIG 510

*Coefficients are shown for all categories of a predictor. Dummy-variable regressions were estimated, and the coefficients transformed by the procedure described in Melichar to restore an effect for the category missing from the regression equation. tSIG means significant at P .05. NS means not significant.

4 Currently employed respondents only. Results are very similar for regressions including nonemployed respondents. (This is accomplished by adding another dummy variable to coworker status for "nonemployed" persons.) Results available from the author.

The negligible impact of kin and coworker status on contact frequency is summarized by comparing these models: Y -f[Neighbor] and Y = [Neighbor, Kin, Coworker]. The increment in R2 from the first to second model is not statistically significant (P - .05) in five of six tests. Specific type of kin does affect contact frequency: Upper-generation and lower-generation kin friends are seen more than average (all coefficients positive, 2-31 days for upper, 1-59 days for lower). Same-generation kin are seen less than average (416 coefficients negative, 1-12 days below average). These effects persist even when a control for distance between

Adult Friendships 1 1305


ego and friend is included. Many of the cross-generation ties are between parents and (adult) children; many same-generation ties are between siblings. Adams (c) cites research that affection is greater for parent-child ties than sibling ties. Also, there are probably more obligations to provide economic and emotional supports between parents and children, than between siblings. These two factors may account for the different levels of contact for types of kin. In summary, the hypotheses that multiplexity boosts contact frequency

is supported for neighbor friends and partially supported for coworker friends. As hypothesized, multiplexity has its strongest effect for neighbor friends. Increased contact with neighbor friends is most likely due to distance (i.e. high opportunities for contact) rather than to added roles or mutual interests that neighbors have.

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

Summary and Conclusions


The incorporation of kin, neighbors, and coworkers into close adult friendships is frequent in the two cities studied. Why does such multiplexity occur? We suggested three explanations: First, neighbors and coworkers are easily accessible. Because face-to-face contact facilitates friendship, people are likely to develop friendships with neighbors and coworkers. Second, important similarities are unique to kin, neighbors, and coworkers. These special interests and experiences held in common may be attractive when adults develop friendships. Third, kin, neighbors, and coworkers tend to offer certain services and emotional supports. When these affiliations are added to a friendship, the tie is more secure and holistic. Examination of social differentials in multiplexity lends support to all three explanations. Multiplexity in kin, neighbor, or coworker affiliation tends to be repeated in the network of close friends, but two or three affiliations seldom appear in the same friendship. Repetition and segregation occur because of (a) restricted opportunities for friendships and (b) positive preferences for developing close ties with one's kin, neighbors, and coworkers. Preferences are probably most important for repetition of multiplex ties. When an individual likes and values contact with kin (neighbors, coworkers), he or she will use that preference when choosing all close friends. Opportunities are also important: The pool of eligible friends differs among individuals, and some pools are especially full of kin, neighbors, or coworkers. Even if individuals choose randomly from such pools (i.e. without preferences), some repetition of multiplex ties will result. Opportunities are probably most important for segregation of kin, neighbor, and coworker statuses in adult friendships. In industrial societies, the pool of eligible friends contains few people with several affiliations

1306 1 Social Forces 1 vol. 57:4, june 1979


(e.g. neighbors who are kin). In fact, such overlap probably appears less often than under conditions of random mixing. When the pool contains so few highly multiplex people, few close friends will be highly multiplex. Preferences also may influence segregation: Highly multiplex ties are very demanding. Despite the promise of diffuse involvement, people may be reluctant to develop highly multiplex ties. The data, however, offer some evidence against this second explanation. Preferences appear to encourage high multiplexity in friendships, rather than to discourage it. In summary, we believe that preferences largely account for repetition of multiplexity, and opportunities for segregation of kin, neighbor, and coworker statuses. These explanations cannot be tested directly with the Detroit and Altneustadt data. Having kin, neighbors, or coworkers as friends must be consequential, influencing the activity, intimacy, durability, and exchanges in a friendship. Virtually none of these consequences has been clearly specified or tested. In this analysis, how often adult friends see each other was examined. Being neighbors greatly increases contact between friends. Being kin or coworkers does not. The former effect is almost certainly due to the high access that neighbors have to each other. The structure, causes, and consequences of multiplexity in friendship dyads and networks are scarcely known. This paper is devoted mainly to structure, with speculation about causes and investigation of one consequence. To understand better why adults acquire multiplex friends requires data tracing friendship development, not simply enumerating friendships at one point in time. To know what benefits people gain from multiplex ties requires data on what adult friends do together and how they rely on each other for goods and services.

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

Notes
1. There may be other social and demographic similarities among kin, neighbors, and coworkers (e.g. neighbors often have similar incomes). These are excluded from the discussion since we want to identify reasons for multiplexity that are specific to being kin, neighbor, and coworker. 2. Verbrugge (b) shows that a powerful basis for friendship choice among adults is similarity in social and demographic characteristics. It is due both to high opportunities that statussimilars meet and to high attraction between them. 3. The items for Detroit: "Now would you think of the three men who are your closest friends and whom you see most often. They can be relatives or non-relatives, as you wish.... For convenience could you give me just their first names?" "Are any of the men you have named relatives of yours?" "Do any of the men live in this neighborhoodsay, within 10 minutes of hereor do they live somewhere else in the Detroit area, or outside of the area?" "Do you see regularly where you work?" The items for Altneustadt (translated from German): "Please think for a moment of the three persons with whom you are close friends and with whom you meet most frequently. Relatives as well as unrelated persons can be so regarded. ... Could you please give me the first name of each of three persons?" "Is any one of the just named persons related to you?" "Who of these friends live in [Altneustadtl, in the near environs of [Altneustadt], that is, perhaps a half hour away, and who lives further [sic] away?" "Do you see usually at your place of work?"

Adult Friendships 1 1307

4. North American studies show that, if these friends are interviewed, 35-45 percent reciprocate the choice and nominate ego as a friend or intimate (Laumann; Shulman). 5. Two other sources of expectations have been suggested: (1) Census data for a population. There are two difficulties with Census data. First, they force one to assume that "people eligible for friendship" are geographically restricted to that population. Second, they lack information on the frequency of kin and coworker affiliations in the population, although they do provide data for "neighbor" indicators. (2) Characteristics of respondents (egos) themselves, taken from the ego marginal distributions. Two difficulties arise. First, this forces the same geographical assumption as for Census data. Second, it ignores the fact that some people are truly more likely to be met and recruited as friends than others (by all types of egos). The choice of the Friends marginal is justified at greater length in Verbrugge (b). 6. Data on intimates ("the persons outside your home that you feel closest to") include relatively more kin (about 40-50% of all nominations), and relatively fewer neighbors and coworkers (Shulman; Weiman, b). Apparently, people often feel close to kin but the kin are not thought of as "close friends." 7. In Altneustadt, some friends are probably more than ten minutes distant from the respondent but are coded as "neighbors." This factor alone may cause higher neighbor choice in Altneustadt than in Detroit. 8. The percent is based on 108 off-diagonal cells, including both odds and marginal ratios. A few of the odds ratios (for coworker) are redundant. 9. It should be noted that ego and a distant friend probably lived near each other in the past. The "choosing" refers to current nomination of best friends, not current formation of friendships with people living far away. 10. In Detroit, the contact item is: "All in all, how often do you usually get together with... (outside of work)?" In Altneustadt, it is: "How frequently do you meet (outside of place of work)?" Interviewers precoded the responses into 6 ordinal categories in Detroit, and 5 in Altneustadt (e.g., "once a week"). For regression analysis, these were rescaled to interval scores, expressed as days per year, by assigning one score to each ordinal category. 11. Consider the reverse causal sequence: Can high contact between friends increase their multiplexity? Since kin status is ascriptive, it is changeless and cannot be influenced by contact. Friends who see each other frequently may decide to move closer to each other. Or they may become coworkers, when one alerts the other to a job opening at his/her worksite. Thus, high contact can increase affect and information among friends, encouraging them to become neighbors or coworkers. Although plausible, this causal sequence is less likely than the one hypothesized in the text.

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

References
Adams, B. N. a:1967. "Interaction Theory and the Social Network." Sociometry 30:64-78. b:1968. Kinship in an Urban Setting. Chicago: Markham. c:1970. "Isolation, Function, and Beyond: American Kinship in the 1960s." Journal of Marriage and the Family 32:575-97. Barnes, J. A. a:1954. "Clans and Committees in a Norwegian Island Parish." Human Relations 7:39-58. b:1972. "Social networks." Module No. 26 in Anthropology. Reading, Mass.: Addison-Wesley. Bell, W., and M. D. Boat. 1957. "Urban Neighborhoods and Informal Social Relations. " American Journal of Sociology 62:391-98. Berscheid, E., and E. H. Walster. 1969. Interpersonal Attraction. Reading, Mass.: Addison-Wesley. Blau, P. M., and O. D. Duncan. 1967. The American Occupational Structure. New York: Wiley. Boissevain, J. 1974. Friends of Friends. New York: St. Martin's. Breton, R. 1964. "Institutional Completeness of Ethnic Communities and the Personal Relations of Immigrante." American Journal of Sociology 70:193-205.

1308 1 Social Forces l vol. 57:4, june 1979

Craven, P., and B. Wellman. 1973. "The Network City." Sociological Inquiry 43:57-88. Cumming, E., and W. E. Henry. 1961. Growing Old: The Process of Disengagement. New York: Basic Books. Festinger, L., S. Schachter, and K. Back. 1950. Social Pressures in Informal Groups. New York: Harper. Fischer, C. S., et al. 1977. Networks and Places. New York: Free Press. Frankenberg, R. J. 1966. Communities in Britain: Social Life in Town and Country. England: Penguin Books. Gluckman, M. 1962. "Les Rites de Passage:' In M. Gluckman (ed.), Essays on the Ritual of Social Relations. Manchester: Manchester University Press. Goodman, L. A. 1969. "How to Ransack Social Mobility Tables and Other Kinds of Cross-Classification Tables." American Journal of Sociology 75:1-40. Kapferer, B. 1969. "Norms and the Manipulation of Relationships in a Work Context." In J. C. Mitchell (ed.), Social Networks in Urban Situations. Manchester: Manchester University Press. Klatzky, S. R. 1971. Patterns of Contact with Relatives. Rose Monograph Series. Washington: American Sociological Association. Laumann, E. O. 1973. Bonds of Pluralism: The Form and Substance of Urban Social Networks. New York: Wiley. Lindzey, G., and D. Byrne. 1969. "Measurement of Social Choice and Interpersonal Attractiveness." In G. Lindzey and E. Aronson (eds.), The Handbook of Social Psychology. Vol. 2. Reading, Mass.: Addison-Wesley. Litwak, E. 1960. "Geographic Mobility and Extended Family Cohesion." American Sociological Review 25:385-94. Litwak, E., and I. Szelenyi. 1969. "Primary Group Structures and Their Functions: Kin, Neighbors, and Friends. " American Sociological Review 34:465- -81. Melichar, E. 1965. "Least Squares Analysis of Economic Survey Data." Proceedings of the Business and Economics Section of the American Statistical Association, 373-85. Mitchell, J. C. 1969. "The Concept and Use of Social Networks." In J. C. Mitchell (ed.), Social Networks in Urban Situations. Manchester: Manchester University Press. Mosteller, F. 1968. "Association and Estimation in Contingency Tables." Journal of the American Statistical Association 63:1-28. Newcomb, T. 1961. The Acquaintance Process. New York: Holt, Rinehart & Winston. Reiss, P. J. 1962. "The Extended Kinship System: Correlates Of and Attitudes On Frequency of Interaction." Marriage and Family Living 24:333-39. Roethlisberger, F. J., and W. J. Dickson. 1939. Management and the Worker. Cambridge: Harvard University Press. Rosow, I. 1967. Social Integration of the Aged. New York: Free Press. Schuman, H. 1973. "Sampling Design for the 1965-1966 Detroit Area Study." In E. O. Laumann. Bonds of Pluralism. New York: Wiley. Shulman, N. 1972. "Urban Social Networks: An Investigation of Personal Networks in an Urban Setting." Unpublished Ph.D. dissertation, University of Toronto. Smith, J., W. H. Form, and G. P. Stone. 1954. "Local Intimacy in a Middle-Sized City." American Journal of Sociology 60:276-84. Srinivas, M. M., and A. Bteille, 1964. "Networks in Indian Social Structure." 64:165-68. Treiman, D. J. 1975. Occupational Prestige in Comparative Perspective. New York: Academic Press. Tyree, A. 1973. "Mobility Ratios and Association in Mobility Tables." Population Studies 27:577-88. Verbrugge, L. M. a:1974. "Adult Friendship Contact: Time Constraints and StatusHomogeneity Effects, Detroit and [Altneustadt], West Germany." Unpublished Ph. D. dissertation, University of Michigan.

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

Adult Friendships 1 1309

b:1977. "The Structure of Adult Friendship Choices." Social Forces 56(December):576-97. Wellman, B. a:1976. "Urban Connections. " Research Paper 84. Toronto: Centre for Urban and Community Studies, University of Toronto. . b:1978. "The Community Question." Toronto: Centre for Urban and Community Studies, University of Toronto. Wellman, B., P. Graven, M. Whitaker, H. Stevens, A. Shorter, S. DuToit, and H. Bakker. 1974. "Community Ties and Support Systems." In L. S. Bourne, R. D. MacKinnon, and J. W. Simmons (eds.), The Form of Cities in Central Canada. Toronto: University of Toronto Press. Wheeldon, P. D. 1969. "The Operabon of Voluntary Associations and Personal Networks in the Political Processes of an Inter-Ethnic Community." In J. D. Mitchell (ed.), Social Networks in Urban Situations. Manchester: Manchester University Press.

Downloaded from http://sf.oxfordjournals.org/ by guest on November 19, 2011

Вам также может понравиться