Persons' attitudes are typically formed in interpersonal environments in which influential positions on issues are in disagreement and liable to change. Our investigation focuses on the formation of attitudes, including shared attitudes and consensus, in groups whose members are communicating their positions on an issue. The communication of an attitude to others may occur in various ways, including spoken or written communication (via an expression of opinion, belief, or preference) and nonverbal communication (via overt behavior, a subtle gesture, or a facial expression). An influential communication, in which one person's attitude affects another's, is an endogenous interpersonal influence, whereas all other effects on attitudes are exogenous influences on the attitudes. The attitudes that are being shaped by these influences may deal with any object about which group members can express a positive or negative evaluation – particular issues, places, persons, events, institutions, symbols, or beliefs. When persons are modifying their attitudes in response to information about the attitudes of other group members, and these other members are doing the same thing, flows of interpersonal influence are generated that permit important indirect effects of members' attitudes on the attitudes of others via intermediaries. In this chapter, we formally present social influence network theory as a mathematical model of endogenous interpersonal influence on attitudes that takes into account the network of such influences.

In this chapter, we will describe a model in which Y(t) affects W(t) at each time t = 1, 2, …. The Y(t) that is involved in this model is an n × n matrix of group members' attitudes toward each other and themselves. Their Y(t) attitudes determine W(t) along with its coupled A(t); and because these attitudes are subject to interpersonal influence, W(t), together with its coupled A(t), generates a Y(t + 1) matrix of attitudes, and so on. From this process, an equilibrium matrix of attitudes, susceptibilities, and interpersonal influences may be produced in the group. The equilibrium W(∞) and A(∞) that may emerge from this process constitute a stable influence network for the group, conditioned on the matrix of initial interpersonal attitudes Y(1). Here, Y(1), in the form of a matrix of interpersonal attitudes, appears as the core construct in determining a group's influence network. The explanation of the origins of stable influence networks in groups has been a longstanding sociological issue, and this model contributes to the theoretical integration of two lines of inquiry related to it – expectation states theory and affect control theory.

Recently, there have been efforts to develop linkages between some of the major lines of work in sociological social psychology – affect control theory, expectation states theory, and social identity theory. Ridgeway and Smith-Lovin (1994) have described possible linkages between affect control theory and expectation states theory. Kalkhoff and Barnum (2000) have described possible linkages between social identity theory and expectation states theory.

In this chapter, we evaluate social influence network theory with data from a series of experiments on small groups: dyads, triads, and tetrads. First, we evaluate whether certain general implications of the standard model, which do not depend on the merits of our measures of A and W, are consistent with the spectrum of observed group outcomes of the attitude change and consensus formation process in groups. Second, we evaluate the accuracy of the model, drawing on subjects' reports of their initial and final (end-of-trial) attitudes on issues, and their allocation of subjective weights to themselves and others in forming their final attitudes. On each issue, we compare the accuracy of this model with the accuracy of a baseline model that predicts the convergence of attitudes to the mean of group member's initial attitudes on an issue. Third, we assess the performance of an optimized model in which group members' susceptibilities are fitted to the data that group members provided about their attitudes and interpersonal influences; and we evaluate the correspondence between the fitted susceptibilities and group members' reported subjective self-weights. Our examination of the latter correspondence is a probe of the phenomenological validity of the model via its derived susceptibility values. Other predictions and applications of the model will be presented in subsequent chapters.

Three Experiments

We begin with a description of the experiments on dyads, triads, and tetrads that we employ to assess the model.

Social influence network theory is based on three constructs – persons' attitudes, susceptibilities, and interpersonal influences. In the previous chapter, we focused on the mechanism underlying attitude changes that result from endogenous interpersonal influences. Under the assumptions of our standard model, this mechanism is y(t + 1) = AWy(t) + (I − A)y(1). Here we focus on the substantive foundations and operationalization of the constructs involved in this mechanism. Theories may be interesting, but to be useful they need to be operationalized. The domain of operationalization includes not only the possible substantive realizations of the theoretical constructs but also the measurement models for these substantive realizations. There is an intimate dance between theory and measurement, and we devote the present chapter to this pas de deux.

The first construct, y(t), defines the substantive domain of our theory, i.e., the n × 1 array of group members' attitudes on an issue at time t. We see the substantive domain of our theory as large. The theory deals with attitudes and their change, but it also encompasses other cognitions that are not easily subsumed by the conventional definition of attitudes as a positive or negative evaluation of an object. Here, we will describe the attitude construct in the general form in which we employ it, and provide several realizations of the construct that are consistent with our specification.

In this chapter, we present an overview of the group dynamics tradition that is our substantive focus, and we present our case for the advancement of this tradition via analysis of the attitude change process that unfolds in interpersonal influence networks. The idea that motivates this book is that some of the important lines of work on attitude change in small groups developed by psychologists (e.g., their work on social comparison, minority–majority factions, group polarization and choice shifts, and group decision schemes on attitudes) may be advanced if a social network perspective is brought to bear on them. In addition, we show how certain lines of current work in sociological social psychology may be advanced with our approach. Sociologists are more likely to pursue these advances than psychologists, given the current emphasis in psychology on social cognition. However, as we emphasize, the influence network and process specified by our theory are a social cognition structure and process. Thus, we seek to move the two orientations into closer theoretical proximity and to build a theoretical interface that speaks to both psychological and sociological social psychologists. By attending to the classic foundations of modern social psychology, to the theoretical perspectives, hypotheses, and findings that constituted the group dynamics tradition, we hope to advance current work on small group social structures and social processes. We revisit the classical past, pursuing an agenda of formal unification, in order to reshape perspectives and trigger new research.

In this chapter, we bring social influence network theory to bear on the explanation of choice shift and group polarization in small groups. A choice shift occurs when, after a group's interaction on an issue, the mean attitude of group members differs from the members' mean initial attitude. Group polarization occurs when the choice shift is in the same direction as the inclination of the mean initial attitude: for example, if on some issue the initial attitude of the average member is positive (negative), then the subsequent attitude of the average member after group discussion will be more positive (negative). An explanation of choice shift is fundamental because it would also explain group polarization. Group polarization always involves a choice shift, but a choice shift can occur that does not entail group polarization (e.g., a choice shift that is in the direction opposite to the initial inclination of the group). Research on choice shifts and group polarization originated with Stoner's (1961) finding on choice dilemmas (issues in which a level of acceptable risk on a course of action is being debated) in which he reported that the decisions of groups involved higher levels of risk-taking than the decisions of individuals. This finding, known as the “risky shift,” stimulated a large number of studies:

Small groups take many forms in natural settings and generally involve discussions of issues of interest to their members. In sociology, at least since Cooley's (1983 [originally published 1902]) observations on primary and secondary groups, various types of groups (families, gangs, communes, work-groups, etc.) have been investigated in field settings as possible sites of important symbolic interactions and the formation of interpersonal agreements. Interpersonal influence networks are implicated in all such groups and, therefore, the development of our understanding of the implications of such networks is a fundamental problem. When group size is scaled upward to the meso-level (e.g., organizations and communities) and to the macro-level (e.g., large populations), influence networks also are present for issues of general salience to the members of these larger collectivities. In such meso and macro collectivities, when interpersonal influences affect members' attitudes toward particular objects and, in turn, their object-related behaviors, influence networks are also fundamental to an understanding of collective action, social diffusion, and behavioral cascades.

Our work, in this volume, focuses on the basic mechanism by which an influence network becomes a meaningful construct, and through which we might analyze the implications of influence networks. We are not exclusively wedded to the experimental settings in which we have examined the implications of influence networks. It appears evident to us that the analysis of interpersonal influences within groups in field settings will be placed on firmer theoretical grounds when it can be demonstrated that particular approaches to the construct of an influence network provide accurate predictions of small group outcomes in experimental settings.

Social influence network theory presents a formalization of the social process of attitude changes that unfold in a network of interpersonal influence (Friedkin 1986, 1991, 1998, 1999, 2001; Friedkin and Cook 1990; Friedkin and Johnsen 1990, 1997, 1999, 2002, 2003). In this book, we bring the theory to bear on lines of research in the domain of small group dynamics that are concerned with changes of group members' positions on an issue, including the formation of a consensus and of settled disagreement, via endogenous interpersonal influences, in which group members are responding to the displayed positions of the members of the group. Newcomb (1951) has suggested, and we agree, that the occurrence of endogenous interpersonal influence is among the basic postulates of social psychological theory:

Social influence network theory advances a dynamic social cognition mechanism, in which individuals are weighing and combining their own and others' positions on an issue in the revision of their own positions. The influence network construct of the theory is the social structure of the endogenous interpersonal influences that are involved in this mechanism.

In this chapter, we examine the implications of interpersonal influences for issues that occur in small groups that are (a) homophilous with respect to a particular sociodemographic characteristic, (b) assembled in a large-scale population, and (c) disjoint in their memberships. These groups are “small” (n = 2, 3, 4, …, 12, … 20, … 100, …, 1,000, …) relative to the size of the population in which they are assembled. We treat these groups as disjoint influence systems with respect to a particular issue (Section 1.3.3), which need not involve direct interpersonal influences or contacts between all pairs of members. The empirical analysis that we present in this chapter addresses the implications of interpersonal influences within small same-sex (i.e., gender-homophilous) groups of dyads, triads, and tetrads for the distribution of issue positions in the population of individuals in which these small groups are embedded.

We focus on a circumstance in which individuals of the same sex randomly assemble into disjoint groups (i.e., groups with no overlapping memberships), enter into their groups with independently formed positions on an issue, proceed to discuss the issue with other members of their groups, and terminate their discussion of an issue after some period of time with either a modified or unmodified position on the issue. Such a set of circumstances describes a common experimental design for the study of small group social processes. These designs have typically been employed to investigate hypotheses concerned with within-group phenomena and, thus far, we have concentrated on such phenomena.