An Integrative Conceptual Framework for Friendship Research

In light of the increase in the number and rigor of studies on adult friendship and the tendency of kin and neighbor relationships to have become more structurally similar to friendship, this is a crucial juncture at which to pause and assess what we know and do not know about adult friendship, to begin a needed theoretical synthesis, to identify gaps in the literature and to produce guidelines for future research. The purpose of this article is to present an integrative conceptual framework, incorporating both sociological and psychological perspectives, for use in these endeavors. The framework posits that the social structural and psychological aspects of individual characteristics operate together to shape behavioral motifs which, in turn, influence friendship patterns (dyadic and network structure and phases). Furthermore, dyadic and network structure and phases affect one another through interactive friendship processes. The elements of this integrative framework and the relationships among them vary by structural and cultural context. Article: In contrast to other forms of intimate relationships in our society, friendship is uniquely voluntary. Whereas relatives are designated by blood or legal ties and neighbors by proximity, friends are selected. Furthermore, friendship is a relatively uninstitutionalized relationship without standard rituals, norms or nomenclature to guide the partners. Yet, friendship choices are not wholly fortuitous, nor is amicable behavior unscripted (Allan, 1989). Scholars who view friendship as voluntary pay special attention to dispositional factors in friendship formation and maintenance. In contrast, those who have a sociological perspective emphasize the effects of social structure and influence largely beyond individual control. These two traditions are distinct in another way as well. Dispositional theorists tend to focus on the interactive processes that take place in friendship dyads, whereas structuralists tend to study the form of individuals' entire friendship networks (Blieszner & Adams, 1992). With two decades of research on record, it is now possible to construct frameworks for the integration of these two perspectives. This is a propitious moment to pause and assess how friend relationships are studied. Kin and neighbor relationships are becoming more similar to the less often studied friend relationship. People now have more freedom in choosing where they live and in determining the quality of their family ties. Understanding how personality and social structure interact to affect friendship should, therefore, provide insight into the dynamics of other, increasingly voluntary, types of relationships. The purpose of this article is to present an integrative conceptual framework in which both sociological and psychological perspectives on friendship are incorporated (see Figure 1). We developed this framework by examining existing friendship research findings, but also by incorporating social psychological, structural and psychological theory in ways that previous friendship researchers have not. The framework can be used to assess what is known about adult friendship and what is not known, to attempt theoretical syntheses of the research, to identify gaps in the literature and to produce a map for future research. The aim of this article, however, is much more modest; it is not intended to be an exhaustive review of the literature (see Blieszner & Adams, 1992; Hays, 1988). Rather, our intention is to stimulate comment that could lead to refinement of the framework, as well as to guide further exploration of the phenomenon of friendship itself. Figure 1 presents a graphic depiction of influences on friendship patterns (see Blieszner & Adams, 1992, for an earlier version of this framework). The basic framework is one that has been used repeatedly in friendship research: individual characteristics (I) affect friendship patterns (II). The elements of this framework and the interactions among them vary by structural and cultural context (III) . As each element of this conceptual framework and the interconnections among them are described in detail below, the distinction between this framework and the typical approach to the study of friendship will become apparent. For now, two departures should be noted. First, friendship researchers have often relied on proxy measures of both independent and dependent variables rather than using direct measures of the concepts of interest to them. For example, many researchers have examined the relationship between sex and frequency of interaction with friends when they were in fact interested in the broader issue of how gender shapes relationships. Although we acknowledge that direct measures of theoretical concepts are sometimes difficult to develop (see Adams, 1989), the approach dictated by our framework involves avoiding the use of mere proxy measures whenever possible. Second, though many friendship researchers have reported findings about the effect of individual characteristics on various aspects of friendship patterns, very few have described the processes by which such effects occur. Duck (1990) compared this to trying to explain the connection between ingredients and dinner without discussing cooking. The approach outlined here involves explications of the processes connecting individual characteristics and friendship patterns as well as descriptions of the interactive processes connecting the structure and phases of friendships. The arrows in the figure indicate both processes that have been shown by research to exist and those that remain to be discovered. Friendship and individual characteristics The majority of friendship researchers have examined the connection between individual characteristics and some aspect of friendship patterns (see Blieszner & Adams, 1992). Structuralists have criticized research on friendship for relying mainly on dispositional assumptions (e.g. that people with similar attitudes have internalized similar norms and thus behave in similar ways) and on explanations that ignore the effect of individual involvement in structured social relationships (Allan, 1989; Wellman, 1983). Conversely, structuralists can be accused of overlooking the influence of affective and cognitive elements in determining intimate relationships. But there is a more serious problem with much of the friendship literature: researchers often fail to develop any theoretical argument, be it dispositional or structural, to support their analyses. They simply add variables such as age or sex and, occasionally, race or class to the research design without stating why they think these variables might have important effects on friendship patterns. They rarely distinguish, either conceptually or empirically, between the effects on friendship of social structural position (Ia) and psychological disposition (Ib). Thus they neither specify the processes by which these effects take place nor examine the interactive effects of social structural position and psychological disposition. The failure to conceptualize and measure the effects of social structural position and psychological disposition independently is particularly problematic in studies of gender and age differences. Although comparing the sexes reveals, for example, that middleaged men tend to have a larger number of friends and middle-aged women tend to have friendships that are more intimate (Fischer & Oliker, 1983), it is not clear why. It is quite different to say that men have more friends than women because they are more likely to occupy positions that put them in contact with potential friends than to say that men have more friends than women because they are more gregarious. Similarly, it is quite different to attribute the greater intimacy of women's friendships to female psychological capacity than to their opportunities to pursue such relationships. Sex is often used as a convenient proxy measure for one of these concepts or both. Researchers typically add sex to a set of independent variables predicting friendship patterns without discussing it conceptually and without including measures of the dispositional and structural consequences of gender (see Fox et al., 1985; Gillespie et al., 1985; Rubin, 1985, for exceptions). Similarly, many theorists have observed that friendship patterns are likely to change as people make life course transitions (e.g. Allan & Adams, 1989) and as they mature (e.g. B. B. Brown, 1990). Just as the variable 'sex' is typically used in lieu of measures of both its social structural effects and psychological implications, in practice, friendship researchers typically use the variable 'age' as a proxy measure for both stage of life course and stage of development without distinguishing between these two aspects of ageing. For example, Weiss & Lowenthal (1975) found that older adults tended to have more complex and multidimensional friendships than middle aged or younger ones. They interpreted the results in light of differing age-related psychological needs and social norms, though they did not measure these needs or normative effects separately. Using the integrative framework (Figure 1) involves measuring the social and psychological aspects of individual characteristics separately and examining how they combine to shape friendship patterns. These elements and the processes connecting them are described later. Individual characteristics such as age, sex, race and class, are conceptualized in both structural terms, as determining opportunities for and constraints on behavior, and in psychological terms, as predicting dispositions. Social structure encompasses the interconnections among social positions (Ia) whose occupants have access to differing levels of power, prestige and wealth and thus different opportunities for and constraints on behaviors that might lead to friendship. Structural effects on behavior include cultural expectations about how people should act; role demands; and the availability, accessibility or appropriateness of spending time in various types of contexts. Structural opportunities for and constraints on friendship-related behavior vary by the social position of the individual in the context. Because it is difficult or impossible for an individual to change her or his social structural position (e.g. to change birth cohort, sex, race or class), a structural theorist would most likely advise changing contexts so that an individual's behavior options would be improved or constraints lessened (e.g. by moving to a new neighborhood or job). Psychological characteristics include personality, motives and personal preferences (Wright, 1989) that affect behavior — that is, psychological disposition (Ib). Some friendship researchers have focused on dispositional differences resulting from socialization and others have emphasized genetic sources of differentiation (e.g. Rushton, 1989). To the person wishing to change her or his behavior, a dispositional theorist would suggest modifying attitudes or other personality traits. Whether the old attitudes or motives derived more from genetically-based predispositions or from socialization, it would be disposition that would have to change, not structural location. As shown in Figure 1, the social structural and psychological dimensions of individual characteristics affect one another through the processes of internalization and interpretation (see Cooley, 1964 and Mead, 1962, for classic discussions of these processes). An individual internalizes social structural expectations and these, in turn, affect her or his disposition. For example, a woman might internalize expectations that women are co-operative and thus be predisposed to this type of interaction. Conversely, an individual's disposition can affect how she or he interprets social structural opportunities and constraints. For example, a shy person might hesitate to attempt to establish friendships across social group boundaries, but a gregarious person might ignore such constraints. Together the social and psychological aspects of an individual's characteristics shape her or his behavioral motif, which becomes the foundation for the friendship patterns in which the person engages. A behavioral motif consists of the constellation of both the routine (Duck, 1994) and unpredictable aspects of an individual's daily activities and her or his responses to them. Individuals do what they are predisposed to do within the structural opportunities and constraints that confront them. A description of an individual's behavioral motif would include the activities in which he or she engages; whether participation in each of them is regular and scheduled and, if so, what the schedule is and, if not, how long he or she typically pursues each of them; the extent to which each of the activities provides social opportunities; and how the individual feels about her or his involvement in the activities. Thus, a person's opportunities for and constraints on participation in various activities and her or his predisposition towards involvement in them determine what the individual actually does — the pattern of her or his daily life. It is this process, the process of living from day-to-day, that shapes friendship patterns. Some friendships are based on routine, repeated, predictable interactions and others are formed after chance meetings. The individual's social and psychological characteristics, however, set the stage for the types of relationships that can emerge. Friendship patterns Friendship patterns (section II of Figure 1) consist of three interacting elements that operate at both dyadic (Ha) and network (IIb) levels of friendship: structure (the form of the ties linking an individual's friends such as the hierarchy and solidarity among them, the similarity of their social positions, the number of friends, the proportion of them who know one another and the pattern or connections among them), phases (the formation, maintenance and dissolution of friendship dyads and of clusters of friends within networks) and interactive processes (the thoughts, feelings and behaviors involved in acting as friends). The arrows in section II of Figure 1 indicate that at both the dyadic and network levels, structure and phases influence one another through interactive processes. Dyads are embedded in networks, and thus their characteristics act upon one another. The potential for processes at one level to affect characteristics at the other level is indicated by the dotted box separating the two levels. Friendship structure Dyadic structure. Power hierarchy, status hierarchy, solidarity and homogeneity reflect the internal structure of friend pairs. The power and status hierarchies are independent, vertical dimensions of relationships (McWilliams & Blumstein, 1991). Power is the `probability that one actor within a social relationship will be in a position to carry out [her or] his own will despite resistance' (Weber, 1947: 152). Status reflects the distance between actors in terms of prestige or moral worth (R. Brown, 1965). Solidarity is the horizontal dimension of internal structure, or the degree of intimacy or closeness between dyad members (R. Brown, 1965). Some researchers treat intimacy as a process variable rather than as a structural one, but it is a fairly widely used measure of the strength of social ties and of social distance (Marsden & Campbell, 1984). Because friendships are typically viewed as intimate, at least to some degree, and egalitarian (i.e. with minimal structure, Thomas, 1987), many researchers have failed to examine these dimensions of internal structure. Friendship researchers who have included a measure of 'degree of intimacy' have found variation on it, even though all of the relationships under study were supposed to be close. Research that addresses power or status within the dyad rather than excluding hierarchical relationships a priori could lead to similarly interesting results. Homogeneity is the similarity of the participants in terms of social positions, such as gender, race, occupational status, ethnicity or age. The internal hierarchy of relationships does not necessarily reflect the relative social positions of the participants in the broader societal context. Individuals who occupy similar structural locations could have different degrees of power or status within the context of their relationship. Network structure. The goal of network analysis is the formal representation of the structure of personal relations within the networks in which dyads are embedded (Feger, 1981; Wellman, 1983). The term 'network' is used in many ways (Milardo, 1992). In this article, network refers to the collection of people that the subject considers as friends. A friendship network constantly changes and overlaps with the friendship networks of other people. To study the structure or interactive processes of personal friendship networks then, one must ask respondents to identify the relevant members of the social landscape considering how they perceive their relationships at the time. Members of an individual's friendship network might or might not know that the individual considers them as part of it. Thus, friendship networks must be studied from the respondent's perspective; the boundaries change and cannot be identified by outside observers. The basic structural characteristics of friendship networks include degree of hierarchy, solidarity and homogeneity, as well as the number of participants (size), the proportion of all possible friendships that exist among members (density) and the patterns of connections among an individual's friends (configuration). Network size can be measured by simply asking how many friends a respondent has, but a more rigorous approach is to operationalize size by counting the number of friends a respondent names. Information on each friendship then can be aggregated into measures of other structural characteristics of the entire network. A compromise method combines a global question about the number of friends with a series of specific questions about a few selected friends. Network density is a measure of 'the extent to which links which could exist among persons do in fact exist' (Mitchell, 1969: 18) and is expressed in terms of a percentage of all possible friendship links actually identified (Kapferer, 1969; Kephart, 1950; Niemeijer, 1973). A concept closely related to network density is configuration, the pattern of interlocking clusters of network members (see Feld, 1981; Laumann, 1973; Peretti, 1976; Salzinger, 1982 for discussions of different ways to characterize network configuration). Network configuration can be quite different in two equally dense networks; in one network, for example, ties might be evenly distributed throughout; in another, two or more clusters of people might be connected by only one mutual friend. Friendship phases Relationships are dynamic, developing and evolving over time. Friendships have beginnings, when partners become acquainted; middles, when solidarity and other features increase, decrease, fluctuate or remain stable; and, sometimes, endings for any of a variety of reasons (see Blieszner & Adams, 1992; Hays, 1988, for reviews of friendship phase literature). The lengths of friendship phases vary between people and situations. Any movement from one friendship phase to another might be deliberate or might occur by chance. Although the language used to describe the trajectory of friendship often implies that friends proceed from one phase to another via planned, conscious mechanisms, individuals differ in the extent to which they consciously employ friendship development strategies. Our purpose in this section is not to specify a unitary definition or a rigid sequence of friendship phases. Rather, we wish to emphasize that understanding friendship patterns is enhanced by recognizing that friendships change over time. Dyadic phases. Friendship formation involves movement from stranger to acquaintance to friendship. The beginning phase of friendship involves identification of or attraction to a potential friend, initial meetings with the potential friend (if a stranger) and getting to know the other and letting the other know oneself. Hays (1984, 1985) described attitudinal and other behavioral changes as partners moved from acquaintanceship to friendship. For example, friendships that became closer over time, compared with those that did not, involved more interaction over a broader range of activities. Closer friendships gradually became more dependent on affection than on frequency of contact. The maintenance phase of friendship is perhaps the most variable period both in terms of the processes that occur and in terms of the degree to which partners consciously attend to the relationship. Friends have many different ways of sustaining their interest in, affection towards and involvement with each other (Rose & Serafica, 1986). From time to time, they may consciously or unconsciously evaluate each other, other friendships, friendship opportunities and relevant social circumstances. Friends might decide, for example, whether to retain the friendship at its current level of solidarity, change to a higher or lower level of involvement, engage in different activities together, dissolve the friendship, display indifference to it or any number of other possibilities (Hays, 1989; Holt, 1982; Murstein et al., 1977). Some friendships enter a dissolution phase and others do not, the end state being the product of a series of other events. Friendships can endure for decades, with the assumption of indefinite existence; some will end abruptly, as with serious disagreement or the death of one partner; and others wither away from benign neglect (Shea et al., 1988). Processes inherent in the dissolution phase concern the ways that individuals end friendships, ranging from direct and explicit declaration of the parting to more indirect ones, such as avoiding the friend or failing to initiate encounters (Rose, 1984; Rusbult, 1987). Network phases. Friendship phases occur not only with respect to dyadic interactions, but also with respect to networks. In the latter case, the formation phase involves the development of new friendship clusters and the integration of individuals and dyads into existing ones. The maintenance phase incorporates the ways that individuals, dyads and clusters sustain the network as it is or change it in some manner. Finally, network dissolution involves the elimination of one or more dyads or clusters or (in rare instances) breaking up of all the friendship bonds in the network. Research on network phases has only just begun. Interactive friendship processes Interactive friendship processes reflect the dynamic aspects of the relationship between dyad members and among network participants. These processes are the covert cognitive and affective responses and the overt behavioral events that occur when people interact (Kelley et al., 1983). In other words, they are 'the adverbial properties of interactions' (Duck & Sants, 1983: 31). Interactive processes in dyads. Cognitive processes reflect the internal thoughts that each partner has about heror himself, the friend and the friendship. These thoughts concern, for example, how one assesses the stability of the friendship, explains shared experiences and interprets one's own behavior and one's partner's intentions or needs, as well as evaluations and judgements of another's attractiveness, character, similarity to the self and so on. Affective processes encompass emotional reactions to friends and friendship. Empathy, affection, trust, loyalty, satisfaction, commitment, joy and contentment are all positive or pleasurable emotions. Indifference, anger, hostility and jealousy are examples of negative or unpleasant ones. Behavioral processes are the action components of friendship. They include communication, such as disclosure of one's thoughts and feelings. Other behavioral processes are displays of affection, social support, resource exchange, co-operation, accommodation to a friend's desires, co-ordination, sharing activities and interests, concealment, manipulation, conflict, competition and the like. The three types of friendship processes interact with each other, such that cognitive processes can result in affective reactions that in turn influence future actions, behaviors can affect thoughts and emotions, and so on. Individuals differ both in the extent to which they employ interactive processes strategically versus assuming a more passive stance, and in the extent to which their dispositions are oriented more toward one type of interactive process than toward others. People can either express thoughts, feelings and actions so their friends are aware of them or keep them hidden. Although studies of various dyadic processes have been conducted (Blieszner & Adams, 1992), the range of interactive processes researched is not comprehensive and the relationships among types of processes have not been examined. Interactive processes in networks. As Simmel (Wolff, 1950) observed, relationships involving more than two people differ from dyads. In a dyad, the withdrawal of one person from the relationship signals its end, so that the friends are fully dependent on one another. In contrast, a network is more stable than a dyad because it continues to exist even if one member departs. A network also imposes constraints on its members and allows individual participants to shift to other members the responsibility for what they have done or not done (Coser, 1977). The cognitive, affective and behavioral processes discussed above for dyads thus operate differently at the network level. For example, a person involved in an isolated dyadic relationship might feel more responsible for helping a friend in need than would a person embedded in a web of friend relationships (Hobfoll & Stokes, 1988). In addition, Simmel suggested that some interactive processes not possible at the dyadic level can occur at the network level. One person could intentionally cause conflict to attain power or other resources, act as a mediator between disputants, or try to enhance her or his position in the group by taking one side of an argument (Coser, 1977). In one of the few studies of friendship network processes (cognitive domain), Hirsch (1979) examined satisfaction with friendship networks among college students. The significant predictors were the students' satisfaction with friendships that involved multiple activities, not having fixed roles in the friendships, having a group feeling when together with friend network members and engaging in a variety of activities with them. Similarly, Stokes (1983) found that the strongest predictors of satisfaction with social support received from network members were having a medium-sized network and having up to seven confidants in the network. This study illustrates the connection between structural and process features of friendship, discussed in further detail below. Proxy measures of interactive processes. Thus far our analysis of processes has focused on specific thoughts, feelings and actions that take place between and among friends. Other process variables have received research attention as well. These include measures of how often and how long interactive processes occur (the length of acquaintance, the frequency of contact, the recency of contact and the average length of each contact). They also include the variety of interactive processes that take place, assessed by multiplexity (the number of different activities in which friends participate together) and directionality (whether friends behave reciprocally or