A Brief Review of Group Cohesion Research

Dianne Kipnes, PhD and Anthony Joyce, PhD

This article outlines how certain basic considerations regarding the phenomenon of group cohesion and conceptual difficulties associated with its measurement have been clarified by investigations spanning the past 40 years. It is not intended as an exhaustive overview of previous or current research on the construct. 

Cohesion has primarily been conceptualized as the members' "attraction to the group." This definition originated in the 1950 Festinger et al. study and was represented in the Gross Cohesion Scale, the most widely used cohesion measure in the literature (Stokes, 1983). Operationalization proved difficult, and arguments ensued regarding theoretical definition, dimensionality, and measurement. Reviews of the literature in the 1980s were unanimous in their concern regarding definition and measurement of the cohesion construct (Burlingame et al., 1984; Drescher et al., 1985; Evans and Jarvis, 1980; Fuhriman and Barlow, 1983; Mudrack, 1989a, 1989b). Research by Budman et al. (1987, 1989, 1993), Evans and Jarvis (1986), MacKenzie (1983, 1984), and Piper et al. (1983, 1984) attempted to reconcile some of the difficulties. Each researcher developed a definition and a related measure of cohesion. Piper et al. (1983) defined cohesion as a "basic bond or uniting force" (p. 95), a unidimensional concept rated by individuals in the group and measured by the Group-Member-Leader Cohesion Scale. Evans and Jarvis (1986) utilized the "attraction to group" definition in the Group Attitude Scale, which assessed members' feelings about the group. The Harvard Group Cohesiveness Scale was developed by Budman and colleagues (1987) to assess "the glue that keeps the group together." MacKenzie (1983, 1984) defined cohesion as the engagement of the group member, including his or her desire for attendance and sense of participation, and the importance of the group to the respondent. These factors were assessed by the Group Climate Questionnaire (1984).

Progress in definition and measurement did not solve all the methodological problems. Questions remained regarding validity. Should cohesion be measured as a group phenomenon only, or is the aggregate of individual ratings representative of the construct? Is cohesion unidimensional or multidimensional? The conundrum is that cohesion represents a property of the group, but without the assessment of individuals it is difficult to evaluate. Evans and Jarvis (1986), MacKenzie (1983, 1984) and Piper et al. (1983, 1984) chose the aggregated individual rating measure of cohesion as the basis for their data, while Budman (1987) chose the overall group measurement of cohesion as assessed by raters. There was no conclusion as to which definition or scale more accurately captured the essence of cohesion in a group. 

In order to develop a consensus on some of the methodological problems, various investigators correlated existing measures of cohesion (Bednar and Kaul, 1978; Bovard, 1951; Eisman, 1959; Evans and Dion, 1991; Gross and Martin, 1952; Hagstrom and Selvin, 1965; Mudrack, 1989b; Ramuz-Nienhuis and van Bergen, 1960). Eisman (1959) suggested that one way to address the problem was to use multiple measures of cohesion. Regardless of the lack of agreement to emerge from these studies, the development of new scales and related definitions, plus the correlation of existing measures of cohesion, resulted in several concepts important for future cohesion research.

More recently, articles have appeared which highlight these issues and attempt to redirect the focus for cohesion research (Bollen and Hoyle, 1990; Cota et al., 1995a, 1995b; Hoyle and Crawford, 1994; Kipnes, 1998; Kivlighan and Lilly, 1997; Marziali et al., 1997; Ostroff, 1993). Some interesting findings have emerged which in some cases support and in others refute previous research.

Cota et al. (1995) reviewed the ongoing controversy regarding the definition and measurement of cohesion. The authors described cohesion as a multidimensional construct with primary and secondary dimensions. Primary cohesion was applicable to all groups, while secondary cohesion was specific to varying types of groups. The authors cautioned that their approach was limited in that it described the content of the construct but not the inter-relationship and development of the various factors that influenced cohesion, such as group developmental level or the use of individual or group level data. Bollen and Hoyle (1990) defined cohesion as consisting of two parts-the individual's sense of belonging in a group and his or her feelings of morale. However, in spite of their attempt to integrate both group and individual level concepts in the definition, the group concept was still defined by individual ratings. 

Recent concerns regarding the measurement of cohesion in group suggest that the unit of analysis should be congruent with the theoretical constructs utilized (Ostroff, 1993; Hoyle and Crawford, 1994). Particular phenomena in group can be investigated by either group and individual level methods, or both levels, but a theoretical basis should accompany the particular level of analysis. The experience of belonging to a group may be best assessed by the group members, while group level phenomena, such as cohesion, may be best represented by an observer-based group measure. 

Results from doctoral work by Kipnes (1998) touches on several of these issues. This research supports the contention that cohesion is a unidimensional construct, which individual and group measures of cohesion are distinct and separate constructs, and that levels of cohesion vary based on whether group or individual assessments are used. Preliminary results suggest that that factor of conflict, including its resolution, may have an important relationship to cohesion. 

Clinicians have long assumed that good group outcome and cohesion are related, but outcome has been as difficult to define and operationalize as cohesion. Consequently, the literature reveals an assortment of outcome measures. Marziali et al. (1997) related curative process factors such as group cohesion (member-member bonds) and group alliance (member-therapist bonds) to outcome. The two processes correlated significantly but independently with outcome. Kivlighan and Lilly (1997) examined the relationships among cohesion, conflict, developmental level of the group, and outcome. Cohesion was found to be related to outcome based on the level of cohesion midway through the group and the pattern of cohesion development. Perhaps more importantly, however, Kivlighan and Lilly (1997) found that cohesion may not be the most important influence on outcome: the concept of avoidance (which they compared to resistance) and conflict have very influential roles in the determination of good outcome. In contrast, Kipnes (1998) found no direct relationship between individual and group level cohesion factors and outcome. 

Recent research on cohesion is illuminating some of the puzzles investigators have struggled with in the past and suggest some alternate investigative directions for researchers and clinicians. The literature continues to present opposing views: cohesion is multidimensional (Cota et al., 1995) or unidimensional (Kipnes, 1998); cohesion is related to outcome (Kivlighan and Lilly, 1997; Marziali et al., 1997) or not (Kipnes, 1998). However, underlying these opposing viewpoints remains an emphasis on several factors which are regarded as essential for future cohesion research: the importance of the various levels in the group and their integrity-the intrapsychic, the interpersonal, and the group as a whole; the development and change of cohesion in the group over time; the influence on cohesion of concepts such as avoidance and conflict; and the interplay of various factors in the group, including cohesion, on outcome. In sum, current views hold cohesion to be a dynamic concept, which both influences and is influenced by numerous factors in the group. Most importantly, perhaps, cohesion can no longer be regarded as only a simple summation of individual assessments at a fixed point in time in the group but is also a phenomenon tapped by group-level assessments, and both individual and group cohesion constructs manifest regularities in the pattern of development over time in the group. 

As can be seen from the foregoing, a construct, which is regularly taken for granted by practicing group clinicians has posed serious challenges for researchers interested in measurement and prediction. Progress over the past 40 years has been made in terms of clarifying multiple issues pertaining to the assessment of cohesion as a phenomenon and as a predictor of group outcome. The next 40 years should indeed be an exciting period in this area of group therapy research. 

Editor's Note: A complete list of the references cited in this article can be obtained by writing or calling the senior author, Dianne Kipnes, PhD; address: 800 Oxford Tower, 10235-101 Street, Edmonton, Alberta, T5J 3G1, Canada; phone: 403-426-6365; fax: 403-429-4778.

This article was published in the December 1998/January 1999 issue of The Group Circle.