representation et works
TRANSCRIPT
-
7/28/2019 Representation Et Works
1/30
What Does Love Mean? Exploring Network Culture in TwoNetwork Settings
King-To Yeung, Rutgers University
AbstractMeaning matters in the way people form social ties. Adopting an unconventional analytictechnique the Galois lattice analysis I show how network researchers can uncoverrelational meanings using conventional research techniques (i.e., closed-ended networksurveys). Galois lattice analysis also inspires new ways of conceptualizing relational meaningsin terms of the duality of persons and relationships, that is, how actors understandings of eachother as persons define the understandings of their relationships with them, and vice versa.The co-constitution of these dualistic meanings thus defines a network culture. Comparingtwo communal settings in which the meaning of love is constructed, I demonstrate thatdifferent network cultures produce different meaning structures that guide how actors relate toone another, resulting in different degrees of group stability.
Social network researchers often speak of networks in terms of their structures, as
configurations of social relationships or ties constructed by social participants. Due to
different methodological constraints (e.g., those of closed-ended survey questions), the
meanings of social ties are often inadequately understood. Frequently, analysts assume that
the meaning of a relationship (e.g., friendship) is self-evident. What people really mean by the
word friend is thereby bracketed as a fixed meaning; meanwhile, analysis goes on to focuson friendship networks as patterns of meaningfully unified ties. Although it is certainly
plausible to think of all network members sharing a common understanding of social
relationships within groups, this assumption is more problematic when we try to address
comparisons across different network settings. In a different group context, friendship may
mean one thing in one group and something else in another. It is therefore necessary forsocial network researchers to examine not only the structural configuration of ties (i.e.,
network properties) but also the meaning structures of different groups. We must ask how
culture affects the ways in which people relate to one another and how these culturally
configured relationships bring forth certain social outcomes (Fine and Kleinman 1983).
Culture exists in many forms. Depending on researchers theoretical orientations, they mayconcentrate on material productions, ideology, beliefs, typifications, practices, and so on (Sewell1999). Focusing on what people really mean by having this or that relationship with some
persons taps into the subjective aspect of culture (Wuthnow and Witten 1988:54). These
meanings and their structures also link culture to cognition (DiMaggio 1997). Recently, network
researchers have shown increased interest in the cognitive aspects of social networks by
examining the difference between the social networks in the minds of network actors and the
This work was partly supported by a grant from the National Science Foundation, SES-99-06452. An
earlier version of the paper was presented at the American Sociological Association meetings in August
2002. I am indebt to John Levi Martin, Paul D. McLean, and two anonymous reviewers for theircomments on early drafts, and to Benjamin Zablocki for his permission of using the Urban
Communes Data. Thanks also to Ann Mische and Karen Cerulo for their inspirations, and to Andrew
Grossman for editorial assistance. Direct correspondence to: King-To Yeung, Sociology, Rutgers
University, 54 Joyce Kilmer Avenue, Piscataway NJ 08854-8045. E-mail: [email protected].
The University of North Carolina Press Social Forces, Volume 84, Number 1, September 2005
-
7/28/2019 Representation Et Works
2/30
network structures observed by researchers (Carley and Krackhardt 1996; Krackhardt 1987a).
The goal of these researchers is not to reduce society into a mental schema, but to understand
the constructural dynamics by which actors subjectivities can generate social structures,
which can in turn shape the action and relationships of the actors (Carley 1986).
In this paper, I follow these lines of inquiry by specifying the linkage among culture, cognitionand social network. Network culture, I propose, is a system of meanings generated by the social
ties of actors (Fine and Kleinman 1983; cf. Eliasoph and Lichterman 2003). Specifically, network
culture interweaves two sets of understandings about actors and their relationships. On the one
hand, we understand each other as persons who possess various personal characteristics, from
personality and body types to skills and knowledge. These personal characteristics are brought
forth in social interaction and hence give meanings to social relationships. On the other hand,the quality of a relationship perceived by actors e.g., whether it is close or tense also
constructs relational meanings. The interconnection between these two sets of understandings
is what I term the duality of persons and relationships in a network culture.
I empirically examine this duality by mapping the interconnection between a set of
personality characteristics (e.g., loving, supportive, dominant, dependent, etc.) and a set of
relationship qualities (e.g., exciting, tense, awkward, hateful, etc.) with a formal method knownas the Galois lattice analysis. As a descriptive tool, Galois lattice analysis has attracted the
interest of many network theorists because it formalizes the theoretical understanding of the
duality, the notion that different analytic elements seem to define and constitute each other
(Breiger 2000; Duquenne 1999; Freeman 1996; Freeman and White 1993; Wasserman and
Faust 1994). Researchers have used Galois lattices to examine the mutual constitutions
between two simultaneous orderings, such as those between organizational helping practices
and the labeling of those who need help (Duquenne, Mohr, and Le Pape 1998; Mohr andDuquenne 1997), between social movement events and activist membership (Mische and
Pattison 2000), or between Supreme Court justices and the issues on which they voted (Breiger
2000). Yet this technique has not been applied to standard (actor-to-actor) social network data.
In this paper, I use Galois lattice technique to examine the interweaving of actors understanding
of persons to whom they relate and the perceived qualities of those relationships.To demonstrate the utility of this particular conceptualization of network culture with the
use of Galois lattices technique, I examine two network cultures wherein the meaning of
love is constructed differently. Although the notion of love is a key for relationship building,
its meaning can vary widely across individuals and social settings. Indeed, the meaning of
love is often attached to cultural processes that shape individuals understanding of theirsocial surroundings (Swidler 2001). Does love mean different things among actors in differenttypes of network settings? If so, how would these different meanings shape the structure of
a social network? Using Zablockis (1980) Urban Communes Data Set, a longitudinal study of
60 urban communes in the United States, I will show how the interlocking patterns of persons
and relationships can explain the variations of group constructions about love . I will answer
the question of how cultural differences between ostensibly similar network settings can
have real social consequences.
Meaning, Social Ties and Network Culture
A Structuralist Understanding of Meaning
Only through language can we characterize social relationships. When we say: my
relationship with X is rewarding, or I like X, rewarding and like become what I call
392 Social Forces Volume 84, Number 1 September 2005
-
7/28/2019 Representation Et Works
3/30
relational concepts, through which we understand and express the meaning of social
relationships. A concept representing a social relation can be multivalent and subject to a
wide range of possible interpretations by actors within a social setting (Pattison 1994). Swidler
(2001), for example, shows that the understanding of love varies among individuals who
actively adopt different cultural resources to make sense of what they think love means.Krackhardt (1987a) illustrates that even seemingly concrete concepts, such as giving advice,
are understood differently, resulting in different cognitive pictures that actors use to navigate
their social worlds. Recently, Bearman and Parigi (2004) find that married men and women,
among other social groups, each talk to different networks when they discuss important
things (e.g., relationships); the meaning of social ties thus depends, in part, on the choice
of important things in interpersonal communication. The very fact that there is rarely acomplete congruence between any egos and alters report about their relationship also
suggests the importance of subjectivity in social relationships (Bernard, Killworth, and Sailer
1982; Casciaro 1998; Freeman and Romney 1987).
One way to examine the subjective meaning of social ties is to map out the interconnection
between various relational concepts when actors use them to talk about relationships. When
an individual refers to another as a friend, she may draw upon and compare other conceptsthat are in opposition to friend (e.g., enemy); alternatively, she may invoke concepts that
are similar but not reducible to friendship (e.g., acquaintance or love). A chain of relational
concepts can thereby specify the meaning of a particular relationship. A story from Hochschild
(1983:44) is useful in illustrating the symbolic connection of relational concepts. In her study of
how people construct their emotions, a confused woman tried to rationalize her relationship
with a potential lover by saying: If he is self-absorbed, then he is unlovable, and if he is
unlovable, which at the moment I believe, then I dontlove him (emphasis added). Self-absorption and lovability may have nothing to do with each other if one is not addressing a
romantic relationship, but in her situation the personality characteristic of self-absorbed
implies unlovable. Connecting these two concepts entails a real network consequence, as
she rejected a potential loving relationship with the man.
The meaning of love (or its opposite) can be found in the way the concept is connectedwith some other relational concepts in a web of signification. This structuralist
understanding of meaning, mainly from the Saussurian tradition, proposes that a concept
does not carry any meaning until it is placed in conjunction with some other concepts (de
Saussure [1916] 1959; Lvi-Strauss 1963).1 Because meanings and the organization of
concepts are viewed as inseparable, structuralist analysis focuses on thesystem of relationsamong concepts (Culler 1975).
When members in a group think and talk about their relationships with one another, theinterconnection of relational concepts used by members constitutes a system of meaning
that is specific to the groups culture. If we could map out the meaning structure in which
relational concepts connect to each other in different network settings, we may be able to
understand how social relationships are rendered meaningful in a particular context. The
structure of relations among concepts generated from a network setting can be considered
as a network culture.
The Duality of Persons and Relationships
What are the components of a network culture? That is, what kind of relational concepts should
be included in a symbolic system of social relationships? An initial answer may mention thoseconcepts related to social ties, such as friendship ties, co-worker ties, and family ties. We say,
we are friends or we are lovers to signify the meaning of a tie. However, these nominal
Exploring Network Culture in Two Network Settings 393
-
7/28/2019 Representation Et Works
4/30
categories frequently fail to tell us the real meaning of a social relationship (Fine and Kleinman
1983). We may better indicate our meaning by evaluating the quality of our relationships through
terms such as our relationship isgreat or this relationship isboring. Relationship qualities can
reveal more about how peoples feelings about their relationships.
Feelings about the quality of a relationship, however, constitute only one side of networkculture. Since social ties are constructed by individuals, we need to take into account the
various characteristics (age, body types, personalities, skills, knowledge, etc.) attached to
network participants when they interact with one another. These individual characteristics can
shape how social ties are formed within a group. In short, the notion of persons is as
important as that of ties in network culture (Casciaro 1998; Emirbayer and Goodwin 1994).
Simmels ([1908] 1971) analysis provided an early insight for integrating persons into socialties. A social circle involves only a selective part of a person or individualities, and each
individuality corresponds only to a particular social context. Consequently, in a quantitative
sense, [i]ndividuality in being and action generally increases to the degree that the social
circle encompassing the individual expands (Simmel, [1908] 1971:252). Breiger (1974)
formalizes Simmels idea as the duality of persons and groups. On the one hand, individuals
comprise a group, and intergroup relationships are determined by the ways people belongto multiple groups; on the other hand, these groups define individuals (multiple) identities
and their interpersonal relationships. Thus, actors and relationships are not irreducible
phenomena; rather, they constitute one another (Breiger 1974:183).2
In order to understand what network culture means to people, one further step may be
taken to treat persons as embodiments of meaningful relationships. For example, McLean
(1998) shows how the meanings of persons and relationships and the interpretation of
those meanings can shape the configuration of ones social networks. Framing socialrelationships for 15th century Florentines meant constructing the appropriate presentation
of the self as a person, and these presentations became a key for defining a social
relationship. Relational concepts such as amicizia (personal relationship) or onore (honor)
invoked culturally appropriate frames for interpreting both persons and relationships.
Consequently, social ties (e.g., patronage) could be built through a dynamic process ofsymbolic signification that also facilitated strategic actions.
Alexanders (1992) work on civil society connects most explicitly individual attributes and
qualities of social relationships. The subjective meaning of civil society, according to
Alexander, originates when people cognitively link personality characteristics with social
relationships that are deemed appropriate in civil society. For example, a person adjectivallycharacterized as civilized is thought to be capable of forming civil relationships. While acivil culture may not emerge as simply as a one-to-one mapping between civilized person
and civility, the connections between characteristics of persons and qualities of social
relationships form an important part of the cultural meaning of civil society.
Thus there seems to be a cognitive link between how we understand each other in terms
of the personality characteristics we attribute to some persons and how these characteristics,
in turn, affect the way we evaluate our relations with them. Conversely, the evaluations of our
relationships with some persons shape how we characterize their personalities. IfA findsB tobe unfriendly, he may also feel that his relationship withB is tense. Dually, ifA perceives his
relationship withB asproductive, he may attribute toB a friendlypersonality.In this paper, I formalize this mutual connectivity and define network culture as the
duality of persons and relationships. Given a set of personality characteristics and a set of
relationship qualities, the meaningful connection (or distinction) between any two relationship
qualities is defined as the number of personality characteristics they both associate (ordissociate). The meaningful connection (or distinction) between any two personality
characteristics is defined, conversely, as the number of relationship qualities associated (or
394 Social Forces Volume 84, Number 1 September 2005
-
7/28/2019 Representation Et Works
5/30
dissociated) with both. Focusing on subjectivities that connect persons to relationships, and
vice versa, may tell us the meaning of social ties and help us better understand the culture
of a social network. To empirically examine this meaning structure, we need a method that
can map out the dual relation by which two sets of analytic elements seem to define one
another. Galois lattice analysis provides such a means, offering a unique perspective on howwe may think about the relationship between persons and relationships in a form of duality.
Galois Lattice Analysis
The technical aspects of Galois lattice analysis related to social sciences have been discussedin Breiger (2000), Duquenne (1995, 1999), Freeman and White (1993), and Wasserman and Faust
(1994). Here I describe how a lattice is read. To construct a Galois lattice we need a matrix from
a two-mode data set showing the binarized relationships between two sets of entities. 3 In this
paper, these two-mode data consist of a set of egos attributions of the personality of alter
(hereafter personality characteristics) and a set of egos perceived qualities of his or her
relationship with alter (hereafterrelationship qualities). In the hypothetical matrix in Figure 1, eachcell represents the presence or absence of an association between a personality characteristic
and a relationship quality. For technical reasons which will be discussed later, I define an existing
association as a statistically positive correlation between one personality characteristic and one
relationship quality at the dyadic level. The presence of an association (coded 1), for example,
between DOMINANT and hateful means that when egos in setting K judge alters as
DOMINANT, egos are also more likely (than chance) to perceive their relationships with alters as
hateful.4
Conceptually, this association implies what structuralists would call signification. Inother words, DOMINANT signifieshateful, and vice versa, in group settingK.
Moving beyond independent relations within each cell, we examine how row elements and
column elements are related in set relations. In Figure 1, there are two sets of elements:P=
{NARCISSISTIC, DECISIVE, SEXY, DOMINANT, DEPENDENT, SUPPORTIVE} andR = {loving,
hateful, jealous}. Each personality characteristic (a row element,p P) or relationship quality (acolumn element,rR) is associated with a set of relationships and personalities, respectively.
For example, jealous is associated with the set {NARCISSISTIC, SEXY, DOMINANT,
DEPENDENT}. DECISIVE correlates with the set {loving, hateful}, which is also a subset of
{loving, hateful, jealous} associated with NARCISSISTIC. This subset relation indicates that
NARCISSISTIC relates to all relationship qualities with which DECISIVE is also associated.The Galois lattice in Figure 1 essentially provides a visual representation of all the subset
relations that are partially ordered.5 Each lattice node (p, r) consists of information about: 1)some row elementsp that are contained in some column elementsr; and 2) some column
elementsrthat are contained in some row elementsp.
Consider the two nodes on the right-hand side of the lattice. The lower node labeled
SUPPORTIVE, DECISIVE, SEXY, NARCISSISTIC,loving indicates that a subset of column
elements {loving} is uniquely (among all relationship qualities) associated with a set of row
elements {SUPPORTIVE, DECISIVE, SEXY, NARCISSISTIC}. This set of personalitycharacteristics also subsumes {DECISIVE, NARCISSISTIC}, a subset of personality
characteristics that is associated with {loving, hateful}. A line will vertically connect any twonodes if they form this subset relation, provided also that the set of relationship qualities
associated with the lower node is a subset of those qualities associated with the upper node.
In our example, {loving} is a subset of {loving, hateful}.
In general, each pair of two nodes connected with a line represents one upper and onelower node. The upper node represents a broader set of relationship qualities than the lower
node, but the lower node represents a broader set of personality characteristics than the
Exploring Network Culture in Two Network Settings 395
-
7/28/2019 Representation Et Works
6/30
upper node. In other words, the narrower set of elements represented in each node is a
subset of the broader set of the like elements represented in the other node to which it isattached by a line.6 Since the subset relations between nodes are transitive, any two nodes
so far as there are lines connecting themcan be read in these ways regardless of thenumber of mediating nodes located between them.
Two nodes on the extreme ends of the lattice should be noted. The top-most node consists
of the universal set of all relationship qualities and the zero set of personality characteristics;
conversely, the bottom-most node consists of the zero set of relationship qualities and the
universal set of personality characteristics. Thus the personality characteristic labeled at the top-
most node (NARCISSISTIC) is associated withallthe relationship qualities. And since none ofthe relationship qualities are associated with all personality characteristics in Figure 1, we only
see the labeling of all personalities, but not relationships.To increase clarity, the lattice is minimally labeled (Figure 2). A personality characteristic is
given a label at the highestpossible point in the lattice, such that the characteristic is
associated with the highest number of relationship qualities compared to the same
characteristic represented in other nodes below it. Hence, NARCISSISTIC is labeled at thevery top of the lattice, where it is associated with the greatest number of relationships
compared to all the lower nodes in which NARCISSISTIC is also found. Similarly, a relationship
396 Social Forces Volume 84, Number 1 September 2005
Relationship Qualities
loving hateful jealousNARCISSISTIC 1 1 1
DECISIVE 1 1 0SEXY 1 0 1
DOMINANT 0 1 1DEPENDENT 0 0 1
Personality
Characteristic
s
SUPPORTIVE 1 0 0
Figure 1. Hypothetical Lattice (Fully-Labeled)
DECISIVE, NARCISSISTIC, DEPENDENTSEXY, SUPPORT, DOMINANT
Set of all personality characteristics
SUPPORTIVE,DECISIVE
SEXY, NARCISSISTICloving
DECISIVE, NARCISSISTICloving, hateful
DOMINANT
DECISIVENARCISSISTIChateful
DEPENDENTDOMINANT, SEXY
NARCISSISTICjealous
SEXYNARCISSISTICjealous, loving
DOMINANTNARCISSISTIChateful, jealous
Set of all relationship qualities
NARCISSISTIC,loving, hateful, jealous
Figure 1: Hypothetical Lattice (Fully-Labeled)
-
7/28/2019 Representation Et Works
7/30
quality is given a label for thelowestpossible point in the diagram, such that the relationship
is correlated with the highest number of personalities compared to the same relationship
quality represented in nodes above it. When a personality and a relationship are labeled at the
same node, they are associated. Minimal-labeling does not change the content of the lattice.
To read Figure 2, one can trace downward those relationship qualities with which apersonality characteristic is associated. For example, NARCISSISTIC is correlated withjealous,
hateful, and loving; SEXY is associated with two qualities, jealous and loving; and
SUPPORTIVE is only connected toloving. We read the ordering of relationship qualities with
a similar logic but from the opposite direction. Tracing upward,hatefulis associated with all
the personality characteristics toward which a line is ascending (i.e., DOMINANT, DECISIVE,and NARCISSISTIC). The relationship quality jealous is connected to DEPENDENT, SEXY,
DOMINANT, and NARCISSISTIC; whileloving is associated with SUPPORTIVE, DECISIVE,SEXY, and NARCISSISTIC.
Lines connecting nodes indicate subset relations among elements. The line between
NARCISSISTIC and SEXY indicates that the relationship qualities associated with SEXY
{loving, jealous} constitute a subset of {jealous, hateful, loving} that is associated with
NARCISSISTIC. Tracing one level downward, we see that the set of relationship qualities
associated with DEPENDENT (i.e., {jealous}) is also a subset of those qualities related toSEXY, DOMINANT, and NARCISSISTIC. Hence, we can say that NARCISSISTIC covers all
personality characteristics in the lattice; SEXY covers both DEPENDENT and SUPPORTIVE;DECISIVE covers SUPPORTIVE; and DOMINANT covers DEPENDENT.
If labeled at the very top of the lattice, a personality characteristic is associated with all the
relationship qualities given in the matrix (i.e., NARCISSISTIC). If labeled at the very bottom, a
personality characteristic is associated with none of the relationship qualities. Conversely, if thevery bottom node in the lattice is labeled by a relationship quality, this quality should be
associated with all personality characteristics given in the matrix. Likewise, the very top node is
Exploring Network Culture in Two Network Settings 397
Figure 2. Hypothetical Lattice (Minimally-Labeled)
SUPPORTIVEloving
DECISIVE
Set of all personality characteristics
hatefulDEPENDENTjealous
SEXYDOMINANT
Set of all relationship qualities
NARCISSISTIC
Figure 2: Hypothetical Lattice (Minimally-Labeled)
-
7/28/2019 Representation Et Works
8/30
labeled by a relationship quality if it is associated with none of the personality characteristics. We
have neither of these cases in Figure 2.
How, then, does Galois lattice analysis produce a structuralist understanding of meaning?
It does so in two ways: through patterns of distinction and elaboration.
Meaning Distinction
Patterns of meaning distinction in the lattice are constructed by the uniquely labeled positions of
some elements vis--vis other elements within mode. For personality characteristics, two or
more such characteristics can relate to each other by means of shared associations with somerelationship qualities; likewise personality characteristics can be distinguished from one another
by what they do not share. Conversely, the connection and distinction among relationship
qualities is defined by their degrees of shared connection with some personality characteristics.
In a formal language, the structure of meaning distinction is located in what Mohr and
Duquenne (1997) call the consensus and dissensus of two elements in a Galois lattice.
Consensuses are located at their meet (theandoperation) of two nodes at their lowerbounds.7 In Figure 2, the consensus between DECISIVE and SEXY is defined by all the
relationship qualities they intersect at their lower bounds, i.e.,loving. The dissensus is defined
by those relationship qualities that are included by DECISIVE but not SEXY, i.e.,hatefulandjealous. Thus, the differential meaning between a DECISIVE person and a SEXY person
depends on whether her loving relationship with others is also perceived as hateful or jealous,
respectively. For another example, tracing upwards we can examine how two relationship
qualities relate to each other by sharing some sets of personality characteristics. Thushatefulandloving share DECISIVE and NARCISSISTIC, yethatefulis distinctly related to DOMINANT,
whileloving is uniquely linked to SUPPORTIVE and SEXY.
In short, meaning distinction is a tool for comparing and distinguishing concepts within-
mode. We understand the meaning of any concept by the way it draws a conceptual
boundary through including and/or excluding some other relational concepts across-mode.
Concept Elaboration
Speech codes, according to Basil Bernstein (1971:123), can be divided into elaborated codesand restricted codes. Elaborated codes invoke a wide range of alternative meanings, tending
to be more abstract, general, and flexible. Because of these structural properties, elaboratedcodes allow actors to communicate and reach across a wider range of social circles. In
contrast, restricted codes tend to be particularistic and narrow, consisting only of a small
range of alternative meanings. Understanding restricted codes depends on local knowledge
among in-group members (Bernstein 1971:78-79). The structural characters of linguistic
codes condition the type of social networks actors can participate, hence linking the structure
of language to the reproduction of social classes (for other empirical applications, see Ansell1997; Bergesen and Jones 1992; Cerulo 1995).
Here I follow Bernsteins general typology and define the elaborateness of a relationalconcept by the extent to which it signifies other concepts across-mode: the more association
one establishes with other concepts across-mode (i.e., concept intension8), the more
elaborated this concept will be. In our hypothetical lattice, the concept intension of
NARCISSISTIC includes {jealous, hateful, loving}, which makes NARCISSISTIC the mostelaborated personality characteristic. Its meaning is more multivalent than DOMINANT and
DECISIVE, which respectively lack the notion of loving and jealous. On the other hand,
398 Social Forces Volume 84, Number 1 September 2005
-
7/28/2019 Representation Et Works
9/30
DEPENDENT and SUPPORTIVE are the least elaborated personalities because they are only
associated with one relationship quality. Regarding relationship qualities,loving andjealous
are equally elaborated by each associating with four personality characteristics, whilehateful,
the least elaborated concept, is associated with only three personality characteristics.9
In terms of concept extension, we focus on the link between concepts within-mode. Forexample, different personality characteristics are more in common with one another at the
lower level of the lattice because they share some common denominators. But as a
characteristic moves up the lattice, fewer and fewer personality characteristics would be
qualified to be on the higher level, which requires the characteristic to be associated with
more unique relationship qualities. This illustrates the property of the extension/intension
duality (Duquenne 1995:265): the higher a personality characteristic is placed (i.e., moreelaborated), the larger set of relationship qualities it is to be included in, while the size of the
set to which this personality characteristic belongs is smaller; the lower a relationship quality
is placed, the larger set of personality characteristics it is to be included in, while the size of
the set that this relationship quality belongs is smaller.
Under this framework of elaboration, concepts are hierarchically ordered by their degrees
of multivalence i.e., the number of meanings attached by a concept. Furthermore, lesselaborated concepts are sometimes nested in other more elaborated concepts. For example,
DEPENDENT is separately nested in SEXY and DOMINANT, and these two latter personalities
are themselves nested in NARCISSISTIC. Being most elaborated, NARCISSISTIC is associated
with the relationship qualities with which all other personality characteristics are associated (but
these characteristics do not have all the qualities that NARCISSISTIC possesses). Here, we can
also think of these nesting relations as relations of elaboration. For example, SEXY elaborates on
what SUPPORTIVE means by associating with an additional relationship quality (jealous);NARCISSISTIC elaborates on what SEXY means by bringing inhateful. We can thus identify
several chains of implication from the lattice by means of examining the nesting relations
among concepts (Duquenne, Mohr, and Le Pape 1998:235). For example: DEPENDENT
DOMINANT NARCISSISTIC; SUPPORTIVE DECISIVE NARCISSISTIC; and so on.
To illustrate the utility of the proposed conceptualization of network culture and theapplicability of Galois lattice analysis as a method to capture this concept, I analyze Zablockis
(1980) Urban Communes Data and examine how subjective meaning of social ties is
constructed differently in two network settings. I show how Galois lattices can help understand
Zablockis counterintuitive finding that a high density of loving bonds tends to destabilize a
group, whereas groups with a low degree of loving density tend to have a longer life span. Thekey to this puzzle, I suggest, lies in the different meaning structures that shape what love couldmean in these network settings.
Urban Communes Data
I analyze the data from the first wave of Zablockis (1980) ongoing longitudinal study of 60
urban commune groups in the United States.10 The first wave (1974-5) of the UrbanCommunes Data yielded a sample of 592 adults who lived full-time in various types of
communal settings.11 This wave contains the most complete network information of 4037dyads on members of 49 communal groups.12 The process of meaning construction appears
to be very important to these groups, because the groups provided not only an alternative
way of living outside the norms of nuclear families but also ideologies to help them impart
meaning to the world (Zablocki 1980:191). They are therefore appropriate sites for examiningthe relations between group structures and meaning constructions.
Network information is drawn from a questionnaire distributed to every member in a
Exploring Network Culture in Two Network Settings 399
-
7/28/2019 Representation Et Works
10/30
group. The questionnaire covers a variety of issues addressing each of the members in the
group, from the question of who cleaned the bathroom to the issue of who had more power
in a particular dyadic relationship. Two parts of the network information are utilized here. First,
each respondent was asked to name one or more group members to whom each of the
following ten personality characteristics might apply:13 supportive, decisive, loving, strong,sexy, dominant, charismatic, passive, dependent and narcissistic.
14
The second part of the data recorded members subjective evaluations of the social
relationships they had with other members. Each respondent (the ego) was asked: How do
you feel about your relationship with him/her (the alter)? For each alter, ego responded to a
list of fifteen relationship qualities; nine of them were used in this analysis:loving, hateful,
exciting, awkward, tense, jealous, exploitative, improving, andsexual.15
For each, a respondentindicated whether yes, sometimes, or no best applied to her relationship with alter. The
answer yes is coded as 1, while sometimes and no as 0.16
Each respondent thereby offers a profile of his or her evaluations about each group
member in terms of 10 personality characteristics and nine relationship qualities. To construct
a Galois lattice for each network setting, I use a 10-by-9 matrix consisting of binary relations
that indicate the association between a personality characteristic and a relationship quality.17
Such associations are computed according to whether or not a personality characteristic was
statistically correlated with a relationship quality. A binary code is assigned such that a 1 is
given to those positive correlations that are statistically significant, and a 0 to non-significant
correlations.18 To binarize negative correlations, I rephrase the element in negative terms for
example, fromloving tonot-loving or from PASSIVE to NOT-PASSIVE, aiming at yielding the
smallest number of nodes generated in the lattice for the purpose of clarity. 19 The errors in
network data are typically correlated and conventional statistical packages might yield biasedresults in statistical tests, but the Quadratic Assignment Procedure is appropriate for
correcting this problem (Krackhardt 1987b). Here, correlations are computed by DAMN
(Martin 1999b), which generalizes the QAP procedure to multiple networks (Martin 1999a). For
lattice drawing, I use ELLA (Martin 1999c), with the help of GLAD (Duquenne 1992) and
KrackPlot version 1.7 (Krackhardt, Lundberg, and Rourke 1993).
The Loving Bond Effect on Group Stability
Among the communal groups under examination, Zablocki (1980) observes that the density ofreciprocated loving ties has a negative effect on group stability. Given that a groups loving density
is defined as the number of loving ties mutually recognized by a dyad in proportion to all possibledyadic ties in a group,20 the higher the loving density the higher the groups disintegration and
turnover rates (Zablocki 1980:165, Figure 4.5). Furthermore, loving density remains significant in
decreasing group stability after controlling for factors such as group ideologies, levels of control
and styles of decision-making, while other factors such as group size and individual characteristics
are found to be poor predictors of group stability (Zablocki 1980:158-160).
If love entails sharing of mutual feelings and affections, this finding seems surprising forsociologists who hold that shared sentiments among members increase social cohesion. This
expectation, however, depends on an assumption that love, whatever it means, alwaysaffects social groups (or individuals) positively across vastly different settings. But if we
assume that networks shape meanings of social ties (Fine and Kleinman 1983:98), then we
must examine the variable meanings of loving ties in different interaction settings and try to
understand how different meaning constructions may be related to varying degrees of groupstability. Here I explore the connection of communal settings and their meaning structures,
as they relate to the duality of persons and relationshipsour definition of network culture.
400 Social Forces Volume 84, Number 1 September 2005
-
7/28/2019 Representation Et Works
11/30
Table 1: Group Types and Life Span
To begin, loving density, as defined above, is computed for each of the 49 groups. Groups
that have a loving density of less than 20 percent out of all possible reciprocated ties are named
cold communes, while the term warm communes designates groups with a loving density
of 60 percent or higher.21 Shown in Table 1, the average life span of cold communes is about 115
months; in contrast, the average life span of warm communes is only 64 months.22
Figure 3: Cold Communes (Below 20 percent reciprocated loving ties; 12 groups; 585 dyads;
average life span = 115.33 months)
Exploring Network Culture in Two Network Settings 401
Group
type(percentof lovingdensity)
All
Groups
Cold
Communes(Below20%)
Neither
Cold norWarm
(20%-59%)
Warm Communes(60% or above)
LeaderResidingpatterns
---- ---- ----All
TypesNon-Charismatic Charismatic/
residentleaders
Charismatic/absenteeleaders
AverageLifeSpan (in
months)
100.94 115.33 105.24 63.75 14.5 107.5 66.5
NumberofGroups
49 12 29 8 2 2 4
NARCISSISTIChateful
not-improvingnot-loving
DOMINANT
Set of all relationship qualities
DEPENDENT(associated with no relationship quality)
Set of all personality characteristics
NOT-SUPPORTIVE
awkward
ealous
LOVING SEXYnot-exploitative
SUPPORTIVECHARISMATIC
improving
exciting
NOT-PASSIVEsexual
tenseexploitative
STRONGloving
DECISIVE
-
7/28/2019 Representation Et Works
12/30
Cold Communes
The Galois lattice in Figure 3 shows a general distinct cleavage that sets apart two
interlocking sets of relational concepts in cold communes (see Appendix for correlation
matrixes and the binarized association matrixes). On the right-hand side of the lattice, thereis a cluster of positive personality characteristics (LOVING, SUPPORTIVE, CHARISMATIC,
DECISIVE, STRONG, SEXY, NOT-PASSIVE) and positive relationship qualities (loving,improving, sexual, not-exploitative). These positive relational concepts are elements
conventionally (or in everyday usage) understood as components that make a relationship
pleasant or harmonious.23 Over the left-hand side are clusters of negative concepts that
conventionally denote something bad about the person or relationship: DOMINANT,NARCISSISTIC, NOT-SUPPORTIVE,hateful, jealous, tense, exploitative, awkward, not-loving,
andnot-improving.
The meaning construction in cold communes thus assumes a clear structure: positive
judgments about the personality of ones dyadic partner tend to be associated with other
positive relationship qualities. In other words, actors draw inference about positive
relationships by means of connecting these relationships with some positive personalitycharacteristics they thought the others have. It is therefore not likely (or not more likely than
chance) to find negative personalities to be associated with some other positive relationships,
or negative relationships with positive personalities.
There are two exceptions to this finding. The relationship exciting straddles positive and
negative spheres, as it is associated not only with positive personality characteristics (DECISIVE,
SUPPORTIVE, CHARISMATIC, LOVING and SEXY) but also with a negative one (DOMINANT).
The node occupied by exciting is a bridging node (cf. Freeman 1996:182) because the removalof it would lead to a complete separation of the two parts of the lattice. In the context of
meaning structure, I define a concept as ambiguous when concepts associated with it are
contradictory; i.e., crossing connections in both positive and negative spheres.24 Another
ambiguous concept is DOMINANT. On the one hand, DOMINANT is associated with negative
relationships asjealous, tense, exploitative; on the other hand, it is also linked to exciting, aconcept that is associated with most positive personality characteristics.
Let us further explore the internal ordering among concepts in terms of conceptual
elaboration and distinction. Regarding elaboration, bothloving as a relationship quality and
LOVING as a personality characteristic are the most elaborated concepts among all positive
ones in cold communes. Aloving relationship invokes judgments about the dyadic partneras STRONG, DECISIVE, SUPPORTIVE, CHARISMATIC, LOVING and SEXY. The personality
characteristic of LOVING is also meaningfully rich by its association withimproving, exciting,
loving andsexual. Indeed, LOVING is an umbrella concept covering most positive personality
characteristics: SUPPORTIVE, CHARISMATIC, DECISIVE and STRONG. But LOVING is also
unique by its attachment to sexual, a relationship quality that none of LOVINGs sub-
concepts have.
We can also find out distinctions among concepts by examining what a concept doesnot
mean compared to the others. Aloving relationship is not significantly associated with anynegative personality characteristics in its upper bound, suggesting that it is quite unlikely (or
not more likely than chance) for members to relate jealous, tense, exploitative, awkward,hateful, not-loving, andnot-improving to aloving relationship. Furthermore, a LOVING person
is clearly distinguished from a NARCISSISTIC person because he or she is associated with
none of the relationship qualities of a LOVING personality. The fact that NARCISSISTIC is the
most elaborated concept among all personality characteristics indicates that the bad guy incold communes is the NARCISSISTIC person who invokes all kinds of negative relationship
qualities.25
402 Social Forces Volume 84, Number 1 September 2005
-
7/28/2019 Representation Et Works
13/30
In sum, being in a loving dyadic relationship means a lot in cold communes, as does
relating to a LOVING person. More importantly, the meaning of being in aloving relationship
and being a LOVING one is unambiguous, as these two concepts distinguish themselves
from all other elements that could contribute to a relationship negatively. If cold communes
tend to be more stable than warm communes as observed by Zablocki, it could be due tothese two facts: 1) a clear and relatively unambiguous meaning structure as a whole dividing
positive aspects of relationship concepts from negative ones, and 2) a high degree of
elaboration in the meaning of love, both as a relationship quality and a personality
characteristic. These two factors imply that not only does love have to mean something
unambiguously, but love has to mean enough in quantity to produce group stability.
Warm Communes
There are different types of warm communes, for groups life spans vary according to
whether or not the notion of charisma is important within a particular kind of commune (see
Table 1). Most of the groups we consider here as warm communes are charismaticcommunes, where members seek collective consensus through submitting their
individuality and autonomy to a group led by a charismatic leader (see Bradley 1987:176-77
for a comparable analysis using the same data set).26 Charismatic leaders are involved in these
groups to varying degrees, in some groups leading only a spiritual existence in the group and
leaving most practical daily business to live-in lieutenants, while in others having direct control
over members by residing in the groups. For those groups with resident charismatic leaders,
the average life span was high (108 months), close to that of the cold communes (115months).27 Since I am focusing on understanding group instability in warm communes, I
exclude charismatic communes with live-in leaders from analysis.
Charismatic groups with absentee leaders have, on average, a life span of 67 months,
about half of that of the cold communes. The Galois lattice in Figure 4 shows that there is a
general decrease in the degree of elaboration in the meaning structure as a whole, reflectedby an overall lower number of nodes in the lattice compared to the cold communes. In fact,
four concepts improving, jealous, sexual, and NARCISSISTIC are not associated with any
other concepts at all. More importantly, the notion of love, either in terms of relationship
quality or personality characteristic, has lost its salience and its capacity to convey social
meaning. In contrast to cold communes, both LOVING andloving in warm communes arerestricted concepts here. This result is not surprising, as warm communes are determined by
a high level (or saturation) ofloving density (mean = .88). Statistically, because there is a lowvariation in theloving ties among members,loving would be less likely to be significantly
correlated with any personality characteristic. A relatively unexpected finding is that the
degree of elaboration of LOVING as a personality characteristic has also fallen, even though
LOVING is not a saturated concept (mean = .17). Thus, the decreasing elaborateness of
love cannot be explained by some artificial statistical consequence but may relate to how
network culture indeed works in absentee communes.If love loses its meaning-organizing power in warm communes, what is its replacement?
Figure 4 clearly shows that charisma has taken such a position, forming the core of meaningstructure in warm communes. Although it is certainly unsurprising to have discovered the
importance of charisma in charismatic communes, it is still puzzling why charismatic
communes without live-in leaders would necessarily produce group instability. Further
examination of these communes meaning structure may offer an answer.In these warm communes, a CHARISMATIC personality structurally synonymous with
SEXY is associated with awkward, tense and exciting; it is also an umbrella concept
Exploring Network Culture in Two Network Settings 403
-
7/28/2019 Representation Et Works
14/30
covering DOMINANT, DECISIVE, STRONG, SUPPORTIVE and NOT-DEPENDENT. Furthermore,CHARISMATIC is also an ambiguous concept capable of invoking both positive and negative
meanings. When a person is perceived as CHARISMATIC, his or her relationship with othersis likely to be evaluated not only positively as exciting but also negatively asawkwardand
tense. This ambiguous character of charisma is different from that of cold communes, where
CHARISMATIC conveys only positive concepts such as improving, exciting, and loving.
Because CHARISMATIC as the umbrella concept organizes the meanings of other social
relations in warm communes, its ambiguity necessarily leads to an unclear meaning structure
for the warm communes as a whole.
Further Comparisons: Chains of Implication
To further understand the logics of meaning construction in the two network settings, I draw
a closer comparison between them by focusing on the way love is constructed vis--vis thenotion of charisma. What does it take for a person to be considered as LOVING or
CHARISMATIC? To answer this question, I examine the covering logic among concepts in a
404 Social Forces Volume 84, Number 1 September 2005
Set of all relationship qualities
improving, jealous, sexual(associated with no personality characteristic)
CHARISMATICSEXY
LOVINGloving
DOMINANTDECISIVE
NARCISSISTIC(associated with no relationship
quality)
Set of all personalitycharacteristics
NOT-PASSIVEhateful
awkward
tense
NOT-SUPPORTIVE
exploitative
STRONGSUPPORTIVE
NOT-DEPENDENTexciting
Figure 4:Warm Charismatic Communes with Absentee Leaders (60 percent or above reciprocated
loving ties; 4 groups; 265 dyads; average life span = 66.5 months)
-
7/28/2019 Representation Et Works
15/30
chain of implication that is, the way one concept is implied by another one in a series of
elaboration (Duquenne, Mohr and Le Pape 1998:235).
So far we know that love means much more in cold communes than in the warm ones,
such that it takes much more effort for a person or a relationship to be considered LOVING orloving. In cold communes, LOVING contains the largest set of relationship qualities (improving,
exciting, loving, sexual), implying that it is most difficult for one in the group to claim a LOVING
personality. Likewise, it is also difficult to achieve aloving relationship, because it is associated
with many (positive) personality characteristics. In warm communes, however, LOVING andloving are immediately and mutually constitutive (Duquenne, Mohr and Le Pape 1998:232),suggesting that the notion of love does not construct other meanings besides itself.
Figure 5 shows all the chains of implication for personality characteristics in both groups.
As I have already implied, LOVING is the most elaborated concept, with the longest chain of
implication in cold communes. The particular logic by which love is elaborated can be seen
in the chain, STRONG
DECISIVE
SUPPORTIVE/CHARISMATIC
LOVING. It reads: allrelationship qualities that are associated with STRONG, DECISIVE, CHARISMATIC, andSUPPORTIVE are associated with a LOVING personality. In other words, being LOVING
implies being all of these personalities. The progression of these concepts along the chain
can also be seen as a process of elaboration. DECISIVE elaborates the idea of a STRONG
personality by correlating with an additional exciting relationship. The personality
CHARISMATIC or SUPPORTIVE is an elaboration of DECISIVE, achieved by addingimproving
into its composition. By associating with a sexual relationship, a LOVING personality
elaborates all the concepts below it, suggesting the key to become a LOVING person is tohave a sexual relationship with alter beyond having an improving, exciting and loving
relationship. (See also Figure 3.)If we were to impose an order on these four concepts in the chain, we may say that they
are distributing along a continuum where, on the one end, we find power-oriented concepts
(or leadership qualities in Bradley [1987]) and, on the other end, egalitarian-oriented
concepts. STRONG is a concept reflecting power, i.e., people can be vertically differentiated bybeing STRONG versus being not STRONG (or weak). In contrast, LOVING may emphasize
egalitarianism (see Swidler 2001). Thus, it seems reasonable that CHARISMATIC and DECISIVE
Exploring Network Culture in Two Network Settings 405
NARCISSISTIC DOMINANT
NOT-
SUPPORTIVE
LOVING SEXY
SUPPORTIVE
CHARISMATIC
DECISIVE
NOT-
PASSIVE
STRONG
Cold Communes
CHARISMATIC
SEXYLOVING
DOMINANT
DECISIVE
STRONG
SUPPORTIVE
NOT-DEPENDENT
NOT-PASSIVE
NOT-
SUPPORTIVE
Warm Communes
DECISIVE
STRONG
Figure 5: Chains of Implication: Personality Characteristics
-
7/28/2019 Representation Et Works
16/30
would fall somewhere in between the two ends of this continuum. In general, it appears that
as a concept becomes more egalitarian-oriented, the concept also tends to be more elaborated.
Contrasting against the chains of implication in warm communes in Figure 5, we confirm
our previous observation that the meaning structure of warm communes tends to lack clarity.
The core chain of implication in the warm communes, STRONG/SUPPORTIVE/NOT-DEPENDENT DOMINANT/DECISIVE CHARISMATIC/SEXY, hardly suggests the particular
progression that exists in the cold communes. At the very least, we expect STRONG and
SUPPORTIVE to be disentangled from each other, since the former entails a stronger
hierarchical notion than the latter. Similarly, we might expect DOMINANT and DECISIVE to
be separated, because they seem to signify different degrees of power-orientation, as
DOMINANT may be a stronger indicator of power or authority.This entanglement between concepts of different power/egalitarian orientation indicates,
once again, that the meaning structure as a whole in the warm communes lacks clarity. No
doubt, this meaning pattern has a certain structural logic (otherwise, it would be
incomprehensible to us), but it is probably not clear enough to allow actors to arrange their
relationships in a meaningful manner.
Figure 6: Chains of Implication: Relationship Qualities
Let us turn to the chains of implication for relationship qualities. In contrast to a clear
difference across the two types of communes shown in Figure 5, the two sets of chains in
Figure 6 appear surprisingly similar. For example, in both warm and cold communes, exciting
and loving are structurally equivalent, as they share the same location vis--vis other
relationship qualities. In cold communes, both exciting andloving elaborate on the linkedquality ofimproving (although the way they elaborate onimproving is different: for exciting,
by adding DOMINANT; forloving, by STRONG). In warm communes, exciting andloving arealso structurally similar, as neither covers any particular relationship quality, even thoughexciting is associated with many more personality characteristics thanloving. (See also Figure
4.) One can conclude that regardless of the type of communes, exciting andloving are similar
in the way they structurally relate to other relationship qualities within groups.The fact that the chains of implication among relationship qualities are quite similar across
the two types of communes suggests that group differentiation is derived from elaborations
406 Social Forces Volume 84, Number 1 September 2005
tenseexploitative
exciting
awkward
loving sexual
improvingealous non-exploitative
hatefulnon-improvingnon-loving
Cold Communes
exciting loving
awkward
hateful
tense
exploitative
Warm Communes
-
7/28/2019 Representation Et Works
17/30
of personality characteristics rather than of relationship qualities. The differences in the
meaning structures between cold and warm communes the former being more elaborated
and clearer, the latter less articulated and ambiguous are organized around two personality
chains of implication: the LOVING and the CHARISMATIC chain, respectively.28 Next, I explain
how these two organizing principles of social relational meaning, together with the clarityand elaboration of the groups meaning structures, may affect group stability.
Discussion
Cold and warm communtes are differentiated by their varying degrees of group love, and thetwo major factors distinguishing their meaning structures are their varying degrees of
meaning elaboration and clarity. Whereas the meaning of dyadic love connects many
relational concepts in cold communes, the same concept in warm communes only signifies
itself (i.e.,loving signifying LOVING). In other words, dyadic love means little when everyone
feels love at the group level. As a group measure, loving density seems to have worked
independently of how dyadic partners would think about their loving relationships asmeaningful. This finding points to a potential disjunction between two levels of social
relationships dyadic love and group love at which members construct relational
understanding about each other.
This disjunction between dyadic and group love occurs most frequently in warm charismatic
groups that promote an ideology of universal love. Members in charismatic communes often
form dyadic relationships, imaginary or real, with a charismatic leader, while leaving relationships
with other members within groups unattended. Rather than perceiving love as a particularlypersonal property, members may conceive of love as a generalized group feature by
universalizing love within the group. They often ritualistically reported we love each other.
From the lattices, we see that love despite its ubiquitous existence in warm communes fails
to link up with other dyadic relational concepts, suggesting that love in warm communes does
not provide much meaning for members at the dyadic level.When generalized love is saturated at the group level, members tend to avoid dyadic
intimacy (Zablocki 1980:170). Furthermore, ritualistically claiming love can obscure real
interpersonal conflicts among members. The unclear meaning structure shown in the warm
communes lattice suggests their members are less equipped to distinguish positive
relationships or persons from negative ones. As love no longer serves as the umbrellaprinciple organizing other relational concepts, and as dyadic love no longer structures groupmeaning, members seem to have lost the social compass by which they could direct
meaningful notions of what a good relationship could be. Failing to produce meanings in
dyadic love thus leads to a meaning-anarchy in warm communes. Group solidarity appears,
then, to depend not so much on the degree of shared sentiment, but on the clarity of a
network culture that helps members sort through relational meanings.
Of course, losing love as the anchor of meaning construction does not automatically lead toa necessary breakdown of meaning clarity. In warm communes, what is puzzling is the way
charisma as it becomes an organizing principle for other relationships generates a vague,
unclear meaning structure. How has individual charisma been expressed in warm communes?As mentioned above, the charismatic leaders of warm communes, who do not reside with
their groups, assume larger-than-life existences, and maintain their charisma through
ideological goals (Zablocki 1980:296; see also Bradley 1987). Often, several communal
groups are organized around one charismatic leader, and members from these differentgroups perceive themselves as part of a single spiritual movement. With a strong, externally
coherent group ideology, members in absentee charismatic communes tend to internalize the
Exploring Network Culture in Two Network Settings 407
-
7/28/2019 Representation Et Works
18/30
authority of the charismatic leader much more strongly than groups with a resident
charismatic leader (Zablocki 1980:298). However, precisely because the leader is absent,
members are less closely supervised and controlled, leading to an organizational concern for
maintaining group integration on a day-to-day basis.
To solve this organizational problem in absentee communes, direct control over dailycommunal business is delegated to one or more live-in lieutenants who serve as the conduit of
authority from the charismatic leader to other members. Individual communication between a
member and the charismatic leader is often mediated by these esteemed lieutenants,
bestowed with the rare position of representing the high leader to the group; for example,
often it is the lieutenants who answer the phone when the charismatic leader calls. 29 Thus, due
to organizational necessity, these communes have placed certain individuals closer to thestatus of the charismatic leaders than the others. However, since no single individual besides
the charismatic leader is supposed to be the CHARISMATIC one, such organizational solutions
to group control can, in turn, create problems to group cohesion.
Even without a resident lieutenant, the process through which members ascend to
authority positions can be problematic. Since none of these groups have successfully
routinized and depersonalized charisma to the extent that charisma can be held by anyincumbent leadership position (Bradley 1987:58), members can struggle to become
CHARISMATIC. But winning the game necessarily puts one into a difficult position, for anyone
who resembles the charismatic leader may be considered as a threat to the absent leaders
authority (Zablocki 1980:299). Thus, it is understandable why a person who is perceived as,
or nearly as, CHARISMATIC as the actual leader will elicit uneasy feelings from other
members, as they simultaneously feel exciting, tense, andawkward(Figure 4). This also
explains why qualities of leadership, such as DOMINANT and DECISIVE, equally conveynegative as well as positive feelings among members. SEXY, on the other hand, shares
similarly ambiguous effects not only because such a personality characteristic is often
attributed to the charismatic leader, but also because sexual energy in warm communes is
highly regulated within the group (Zablocki 1980:172).30
In sum, the high degree of instability in warm communes can be understood, first, by theincapacity of members to generate dyadic meanings of love, a source of interpersonal
intimacy. Second, warm communes also lack a clear meaning structure that allows members
to differentiate positive relational concepts from negative ones. The source of ambiguity in
the meaning structure, I suggest, stems from the competition over charismatic qualities
between the absentee charismatic leader and the charismatic ones who live in the group.Since the core organizing principle for meaning construction (CHARISMATIC) is itself
ambiguous, it is therefore not surprising to find that the meaning structure of warmcommunes as a whole falls into a state of meaning-anarchy. Clarity in the meaning structure
of dyadic relationship helps members to distinguish the qualities of one relationship from the
qualities of another; furthermore, clear meaning structure may form thestructural base for
members to build group cohesion at a collective level. Further research will need to focus on
the precise social mechanisms through which an unclear meaning structure can lead to group
instability. This paper has so far only identified two important dimensions of group stability,the degrees of elaboration and distinction in a network culture (cf. Carley 1991).
Conclusion
This paper begins with the simple assumption that meaning matters in the way people formsocial ties. Network survey techniques that use closed-ended questionnaires to derive network
structures often neglect the semantic aspect of social networks. But if studying the meanings
408 Social Forces Volume 84, Number 1 September 2005
-
7/28/2019 Representation Et Works
19/30
of social ties helps to explain structural outcomes of social networks, then analysts must find
ways to more subtly analyze what network participants mean when they engage in
constructing social ties with one another. Adopting the technique of Galois lattice analysis, I
have shown how network researchers can uncover such relational meanings without changing
their conventional research techniques (i.e., the use of network survey). Furthermore, I haveillustrated that the meaning structures uncovered by Galois lattices help explain why some
group settings produce a high degree of group stability while others do not.
As with many other methodological tools, Galois lattice analysis does not operate in
isolation from researchers conceptual and theoretical tasks. If we adopt a structuralist
perspective of meaning and assert that social meanings arise by actors efforts to connect a
(symbolic) concept with some others, then researchers can map out all those conceptualconnections, and investigate the different meaning structures specific to some network
settings. Galois lattices propose and specify an additional way to view these conceptual
connections, in terms of a duality of two sets of analytic elements.
These methodological and theoretical concerns lead to a new conceptualization of
relational meaning. Given a set of egos reports on the personality characteristics of alter,
and a set of egos self-evaluations of the quality of his or her relationship with alter, themeaning structure of a social network is defined by the duality of these two sets of elements.
Simply put, how we understand each other as persons reciprocally defines the way we
understand the quality of our relationships with these persons. At the group level, the
cumulative structure of these dualistic meanings between persons and relationships
constitute the shape of a network culture.
Relationships and love are certainly complicated issues. With the help of Galois lattices,
we can formalize our theoretical understanding of network culture and objectively measurethe otherwise esoteric concept of cultural meanings. Lattices also produce a birds eye view
of how actors draw inferences about each other (qua persons) by means of the types of
relationships they share, and vice versa. More importantly, this birds eye view is often
accessible to neither the network participants themselves nor the researchers who read from
the correlation matrices (such as those in the Appendix). The descriptive power of Galoislattices thus lies in their ability to formalize and simplify two-mode data, clearly specifying how
each mode of the data should be related. Of course, it also means that if researchers cannot
theoretically justify their use of a duality, Galois lattices would not help improve their analytic
power. Essentially, a new method can often inspire researchers to develop new conceptual
frameworks, but no method should be used without researchers careful examination of itsassumptions and theoretical appropriateness.
I have demonstrated that the use of Galois lattices indeed helps us understand the
variations of group stability in two different network settings. I have compared the network
culture of groups that have very dense loving ties (where most paired persons report they
have loving relationships with one another) against that of groups with sparse loving ties.
Previous research has found that groups with a high density of loving ties actually have
shorter life spans than those that do not, which might strike many as counter-intuitive. Butcomparative Galois lattice analysis shows that the two types of groups have different sets of
meanings of relationships. In general, love fosters many (elaborated) positive meanings
between two people in groups with sparse loving ties. Furthermore, the meaning structure ofthese groups is also clearer than that of the seemingly love-full groups, implying that
memberships in the latter groups lack the possibility of arranging their relationships in a
meaningful manner. Groups that can generate a high degree of elaboration andclarity inrelational meanings may be more stable than those that cannot.
This finding, however, remains tentative because of the very technique I am using here.
For example, Galois lattices require a fixed set of concepts that may or may not exist in the
Exploring Network Culture in Two Network Settings 409
-
7/28/2019 Representation Et Works
20/30
actual interactions among members. Particularly, the use of a closed-ended network survey
restricts other potential concepts that members might use to refer to a loving relationship or
a loving person. Further research may adopt an open-ended instrument so that respondents
can generate concepts by themselves (see Bearman and Parigi 2004). Because Galois lattice
technique also requires several steps of data reduction in order to generate a comprehensiblestructure for interpretations, lattices tend to be temporally static or simplified. This may
trouble analysts of culture because a simplified structure can compromise the view that
culture is dynamic, comprised of meanings constructed in particular situations and in on-
going interaction (Eliasoph and Lichterman 2003).
Despite these limitations, this paper shows the importance for network analysts to
address the links among network properties, cultural elements, and network outcomes. If thepatterning of ties shapes the meaning structure of a social network, then analysts cannot
reduce a social tie to a merely structural property. Conversely, we may not be able to
understand tie-structures without attending to how actors construct meanings in their social
relationships. Network culture as proposed in this paper, therefore, serves as an important
theoretical and methodological link, connecting network properties to network outcomes.
Notes
1. Here I am adopting a rather liberal use of the term structuralism of which the Saussurian
tradition which treats the relationship between signs and their referents as arbitrary is
only one branch (see Mohr 1998).
2. For recent uses of the notion of duality, see Breiger (2000); Breiger and Pattison (1986);
Giddens (1979: 81-85); Mohr and Duquenne (1997:310-12); Padgett (2001:223); Sewell
(1992).
3. A mode is a distinct set of entities on which the structural variables are measured(Wasserman and Faust 1994:29). Here we have personality characteristics and
relationship qualities.
4. Although the analysis is conducted at the dyadic level, the resulting correlation
coefficients (or associations) are properties of a network setting.
5. A partially ordered set is a non-empty set with the binary relation , such that for allelements in the set {a, b, c, } satisfy all three conditions: 1) reflexive, such that a a for
any elementa; 2) anti-symmetric, such that ifa b andb a, thena =b for any two
elementsa and b; and 3) transitive, such that ifa b andb c, thena c for all elementa,b, and c. A partially ordered set, however, does not require thata b ora b for a paira,b in a set {a, b, c, } as it is required in a linearly ordered set.
6. Formally, if some set of row elementspi contains all the row element of somepj, we say
pjpi ; the same is true for any two sets of column elements. In each node (ri, pj), thebinary relation is defined as (rk, pl) (ri, pj) ifrkriandpjpl. A line will descend
from (ri,pj) to (rk, pl) if (rk, pl) (ri, pj); equivalently:rkriandpjpl.
7. An elementp Pis an upper (lower) bound forA if and only if for alla A, a p (a p).The meet of two nodes is also the intersection of the sets of elements (within mode)
related to the two nodes.
410 Social Forces Volume 84, Number 1 September 2005
-
7/28/2019 Representation Et Works
21/30
8. Concept intension is the list of properties or attributes associated with one concept.
9. It should be noted that the degree of elaboration of a concept only refers to the within-
mode comparison among concepts. For example, the elaborateness of a personality
attribute is determined by the number of relationships contained in that attribute inreference to some other personality attributes in the system. Across-mode comparison
does not render such comparisons meaningful, especially in a system where the given
number of concepts is different in each mode (e.g., 10 personality attributes and 9
relationship characterizations in our case).
10. The data have recently been made public (see Martin, Yeung, and Zablocki 2001).Descriptions of the data are available at http://sociology.rutgers.edu/UCDS/UCDS.htm
11. The types of communal groups range from religious (Christian and Eastern religions), political,
countercultural to groups that were oriented to personal growth, and to less structured
groups based on friendship. For sampling procedures, see Zablocki (1980:368-74).
12. These are groups that have more or less complete network information. Some of the 60
groups have been omitted from this analysis because: 1) three groups had refused
collectively to participate in the network survey; 2) not all individuals had completed the
questionnaire; since this analysis is concerned with the network culture, it does not make
sense to include groups that have high number of missing dyads. Hence, I exclude 6
groups that have over 50 percent of missing dyads from the analysis.
13. These characteristics are listed in the questionnaire in this order. More than ten personality
characteristics were presented in the survey. Other characteristics such as interested in
kids and interested in fixing up the house were excluded from this study because they are
highly skewed by gender. Hence it is impossible to distinguish whether the response
actually referred to alters personality or alters gender. Other characteristics influential,has a clear vision of the communes future, intuitive, holy, good at resolving emotional
problems, can translate ideas into action, and can motivate me to do more are also
excluded because they were presented in the survey only for a portion of the respondents.
14. There is no limitation to the number of names a respondent could list; hence, arespondent might list no one or everyone in the group. When everyone was writtenunder a category, it was not coded, because we are concerned with personality
characteristics that can be attributed to particular persons.
15. Items such as work together, mind kids together, sleep together, spending free
time together, and agree on communal policy, were excluded because they are not
abstract relational concepts that express a feeling toward a relationship. Other items suchas parental, confide to each other, close, unreciprocated, feel estranged, and
important to me are excluded because they were not presented to all respondents.
16. This coding reduces the frequencies of the positive responses for any relationship or
personality. Hence, the coding scheme allows us to eliminate hesitant responses and
reduce the complexity of the lattice structures (see Breiger 2000 for a discussion).
17. Theoretically, there should be 4037 matrixes, one for each dyad. Not only is it impossible
to compare more than four thousand matrixes, studying each matrix alone can lead to
Exploring Network Culture in Two Network Settings 411
-
7/28/2019 Representation Et Works
22/30
another problem. For instance, if ego denies anyloving feelings toward alter, the column
ofloving in this particular matrix would be 0 across all personality characteristics. Hence,
the relationship betweenloving and dominant and betweenloving and dependent cannot
be independently derived, for they all depend on whetherloving is reported.
18. It should be noted that the magnitude of the correlation coefficients is relatively low. To be
sure, there are many other factors that can explain why ego feels this or that way about alter
and their relationship. Egos view on alters personality may be explained at least by two
other factors: 1) the actual personality of alter (the objective thesis) and 2) the relationship of
ego with a third person (the balance thesis). The goal of this paper, however, is not to explain
the variations across variables, but to show that how relationship characterizations andpersonality attributes matter in shaping network culture. Hence, the use of correlation
coefficients helps demonstrate the descriptive patterns, rather than provide explanation.
19. Chi-square tests for independence yield a similar result and did not change the skeletal
form of the lattice (see Brieger 2000 for discussion on methods of binarization). Zero
correlation coefficients were presented because they are easily interpreted, especiallywhen the difference between positive and negative associations becomes important later.
Statistical significance is deemed to be a better way of binarization. Although we can use
the magnitude of the correlation coefficient above a certain threshold to determine
whether an association is present between two relational concepts, this method is
relatively arbitrary and requires us to assume that all relational concepts are equally
important across all groups. If we impose different thresholds to different pairs of concepts
in different groups we have to assume knowledge that we do not yet have. But since themagnitude of correlations does matter, I code low coefficients (i.e., r < .05) as 0 even when
it is statistically significant. Among all 180 coefficients, there is only one case of this. The
skeletal structure of the lattices does not change by including or excluding this coefficient.
20. I use this definition of loving density throughout this paper. The loving tie here is fromthe same variable used to construct the lattices; i.e., whether or not dyadic partners
characterize the quality of their relationships as loving. The number of all possible ties in
a group is measured by [N (N - 1)] / 2, whereN is the number of members who had valid
data. A valid dyad is restricted only to those dyads with valid data for both respondents.
21. Admittedly, the cut-off points of these groups have an arbitrary character but arereasonable according to Zablockis finding (1980:165, Figure 4.5):
Loving Density Disintegration Rates Turnover Rates
0-20% 0% 28%
21-40% 0% 50%
41%-60% 8% 55%61-80% 33% 63%
81-100% 60% 65%
Member turnover rates almost double when groups exceed a loving density over 20
percent, even though group disintegration rates remain unchanged. Group disintegration
rates also drastically increase when groups attain a loving density level over 60 percent.The cut-off points used in this paper, under 20 percent for cold communes and over 60
percent for warm ones, seem to be reasonable to capture the effects of group instability,
a concept which multi-dimensionally combines both turnout rates and disintegration
412 Social Forces Volume 84, Number 1 September 2005
-
7/28/2019 Representation Et Works
23/30
rates. In this paper, I only highlight the life span of the groups, which turns out to be quite
consistent with other aspects of group stability (see Bradley 1987:175).
22. In order to maximize the clarity of the analysis, I have ignored groups whose loving
densities lie between 20 to 59 percent. Table 1 shows that this middle group has anaverage life span between that of the warm and cold communes.
23. A similar approach to understanding the meaning of concepts in lattices by external
criteria is found in Duquenne, Mohr, and Le Pape (1998:230), who distinguish the positive
and negative concepts by the conventional moral ordering of the concepts.
24. While elaborated concepts contain multiple meanings, so long as they do not embrace
contradictory concepts, they remain unambiguous. Hence, an elaborated concept can
either be ambiguous or unambiguous.
25. We will further explore below why it is not a NARCISSISTIC personality that organizes the
network culture of cold communes, despite the fact that there are more relationshipqualities associating with NARCISSISTIC than with LOVING (see note 28).
26. I ignore warm, non-charismatic warm communes because the total number of dyads is
too small for any meaningful comparison.
27. One reason why resident charismatic groups tend to be more stable is that a resident
leader may channel the saturated loving ties in the group into positive feelings and personalrenewal: the high density of love is tolerable because the love relations themselves are
experienced as renewed [experience] (Zablocki 1980:182, 184-5; see also Bradley 1987).
28. Figure 5 reveals that although NARCISSISTIC is the most elaborated personality characteristic
in cold communes (Figure 3), it only organizes other characteristics in a limited sensecompared to LOVING. NARCISSISTIC is therefore an elaborated concept but not an
organizing principle for the cold communes. Negative relationship qualities in the two kinds
of communes also share similar positions vis--vis other relationship qualities within each
diagram in Figure 6, with tense being a more elaborated concept andhatefula restricted one.
29. I thank Anna Looney, who conducted in-depth interviews with these commune members,for this information.
30. The lattice in Figure 4 shows that in warm communes the meaning of sexual as a
relationship quality is absent. This can be interpreted to mean that either that warm
communes suppress sexual activities among members or that sexual relationships are so
widely permitted among members that they become meaningless. This observation isconfirmed by checking the ethnographic records of these groups. Both eastern religious
groups and free love cults are categorized as warm communes. In the eastern religious
groups, some members are strictly celibate; some marriages are arranged by thecharismatic leader as spiritual marriages, and sexual practices after marriage are
influenced by group doctrine (e.g., regarding the proper preparation for and frequency of
sexual intercourse). In the free love cults, an equal amount of group teaching on sexualpractices exists for the opposite purpose, as members are encouraged to sexually
experiment with other group members. Having multiple sexual partners is not
stigmatized. However, the cost to relationship harmony and stability seems high in both.
Exploring Network Culture in Two Network Settings 413
-
7/28/2019 Representation Et Works
24/30
References
Alexander, Jeffrey C. 1992. Citizen and Enemy as Symbolic Classification: On the Polarizing Discourse ofCivil Society. Pp. 289-308 in Cultivating Differences: Symbolic Boundaries and the Making ofInequality
, edited by M. Lamont and M. Fournier. University of Chicago Press.
Ansell, Christopher K. 1997. Symbolic Networks: The Realignment of the French Working Class, 1887-1894.American Journal of Sociology103: 359-90.
Bearman, Peter, and Paolo Parigi. 2004. Cloning Headless Frogs and Other Important Matters: ConversationTopics and Network Structure.Social Forces 83: 535-57.
Bergesen, Albert J., and A. Jones. 1992. Decoding the Syntax of Modern Dance. Pp. 169-84 inVocabularies of Public Life, edited by R. Wuthnow. Routledge.
Bernard, H. Russell, Peter D. Killworth and Lee Sailer. 1982. Information Accuracy in Social Network Data: IV. A
Comparison of Clique-Level Structure and Behavioral and Cognitive Data.Social Networks 11: 30-66.
Bernstein, Basil. 1971. Class, Codes and Control: Theoretical Studies Towards a Sociology of Language.Schocken Books.
Bradley, Raymond Trevor. 1987. Charisma and Social Structure: A Study of Love and Power, Wholeness andTransformation. Paragon House.
Breiger, Ronald L. 1974. The Duality of Persons and Groups.Social Forces 53: 181-90.
_______. 2000. A Tool Kit for Practice Theory.Poetics 27: 91-115.
Breiger, Ronald L., and Philippa E. Pattison. 1986. Cumulated Social Roles: The Duality of Persons and TheirAlgebras.Social Networks 8: 215-56.
Carley, Kathleen. 1986. Knowledge Acquisition as a Social Phenomenon.Instructional Science 14: 381-438.
_______. 1991. A Theory of Group Stability.American Sociological Review 56: 331-54.
Carley, Kathleen, and David Krackhardt. 1996. Cognitive Inconsistencies and Non-Systematic Friendship.Social Networks 18: 1-27.
Casciaro, Tiziana. 1998. Seeing Things Clearly: Social Structure, Personality, and Accuracy in Social NetworkPerception.Social Networks 20: 331-51.
Cerulo, Karen. 1995.Identity Designs: The Sights and Sounds of a Nation. ASA Rose Monograph Series.Rutgers University Press.
Culler, Jonathan. 1975.Structuralist Poetics. London: Routledge & Kegan Paul.
de Saussure, Ferdinand. [1916] 1959. Course in General Linguistics. Translated by W. Baskin. PhilosophicalLibrary.
DiMaggio, Paul. 1997. Culture and Cognition.Annual Review of Sociology23: 263-87.
Duquenne, Vincent. 1992. GLAD: General Lattice Analysis & Design. A FORTRAN Program for a Glad User.Maison des Sciences de lHomme.
_______. 1995. Models of Possessions and Lattice Analysis.Social Science Information 34: 253-67.
_______. 1999. Lattical Structures in Data Analysis.