Group Dynamics: Theory, Research, and Practice 2001, Vol. 5, No. 3, 220-234
Copyright 2001 by the Educational Publishing Foundation 1089-2699/01/S5.00 DOI: 10.1037//1089-2699.5.3.220
Does Group Climate Mediate the Group Leadership-Group Member Outcome Relationship? A Test of Yalom's Hypotheses About Leadership Priorities Dennis M. Kivlighan Jr. and Jeffrey M. Tarrant University of Missouri—Columbia I. D. Yalom's (1995) hypothesis that group climate mediates the relationship between leadership and outcome was tested. Group leaders (N = 43) recorded intentions, and adolescent group members (N = 233) rated climate after 8 semistructured group sessions. Members also rated satisfaction at termination. Leader intentions comprised 4 dimensions: therapeutic work, safe environment, interpersonal, and group structure. Changing group climates did mediate between these intentions and member outcome. Therapeutic work was negatively related and safe environment was positively related to an increasingly active and engaged climate, which was related to treatment benefit. Group structure and possibly interpersonal intentions were related to a climate decreasing in conflict and distance, which was related to a positive leader relationship. The importance of leaders focusing on group process rather than individual change is highlighted.
For Yalom (1995), the group leader's primary responsibility is the creation of a therapeutic group climate. "If it is the group mem-
bers who, in their interaction, set in motion the many therapeutic factors, then it is the group therapist's task to create a culture maximally conducive to effective group interaction" (Yalom, pp. 109-110). Yalom's writings suggest that the group members have a direct influence on member outcome whereas the group leader's influence is indirect, acting through the creation of a therapeutic group climate, denned as a series of interactional dimensions (MacKenzie, 1983). Group climate should mediate the relationship between group leader activity and group member outcome. Viewing group climate as a mediator provides a framework for summarizing the group leadership and group climate literatures. Baron and Kenny (1986) defined a mediator as a variable that explains the relationship between a
Dennis M. Kivlighan Jr. and Jeffrey M. Tarrant, Department of Educational and Counseling Psychology, University of Missouri—Columbia. Correspondence concerning this article should be addressed to Dennis M. Kivlighan Jr., Department of Educational and Counseling Psychology, University of Missouri, 6 Hill Hali, Columbia, Missouri 65211. Electronic mail may be sent to
[email protected].
predictor variable and a criterion variable. To demonstrate that group climate mediates the relationship between group leadership and group member outcome, three conditions must be met (Baron & Kenny): (a) group leadership must be related to group member outcome, (b) group leadership and group climate must be related, and (c) group climate must predict group member outcome. Is Group Leadership Related to Group Member Outcome? On the basis of a review of 135 studies, Dies (1994) found that leader structure, especially meaning attribution (e.g., Richardsen & Piper, 1986), shows a consistent positive relationship to group member outcome. Meaning attribution involves providing concepts for the members to use in understanding their individual experience or group events. A positive relationship with the leader is also related to group member outcome. Dies concluded "that group members favor and seem to benefit more from a positive style of intervention, and that as leaders become more actively negative, they increase the possibility that participants will not only be dissatisfied, but also potentially harmed by the group experience" (p. 139).
220
MEDIATIONAL ROLE OF GROUP CLIMATE
Coyne, Harvill, Morganett, Morran, and Hulse-Killacky (1990) suggested that group researchers have ignored "the fact that leadership is largely a cognitive process" (p. 33). Research on the covert aspects of group leadership has focused on the cognitive schemas therapists use to assess group processes (Kivlighan & Quigley, 1991) or to decide on specific interventions (Hines, Stockton, & Morran, 1995). In individual treatment, however, researchers have focused on the intentional aspect of the therapist's cognitive activity. Intentions "refer to why, whereas interventions and techniques refer to what the therapist does" (Hill & O'Grady, 1985, p. 3, italics in original). Therapist intentions are better predictors of session climate and client outcome than are overt therapist behaviors (Hill et al., 1988). Stiles et al. (1996) developed the therapist session intentions form (TSI), a session-level measure of therapist intentions. A simultaneous components analysis of the TSI revealed the following seven factors common to psychodynamic-interpersonal and cognitivebehavioral treatments: Treatment Context Focus, Session Structure Focus, Affect Focus, Obstacles Focus, Encouraging Change Focus, Behavior Focus, and Cognition-Insight Focus. Because session intentions have not been examined in group counseling, our initial goal was to identify the salient dimension of group leader session intentions. Because structuring and relationship dimensions are prominent in leader research, we hypothesized that the group leader session intentions would contain at least these two dimensions. In addition, we hypothesized that higher levels of structuring and positive relationship would be associated with better outcome. Is Group Climate Related to Group Member Outcome? Cohesiveness, defined as "the condition of members feeling warmth and comfort in the group, feeling they belong, valuing the group and feeling, in turn, that they are valued and unconditionally accepted and supported by the other members" (Yalom, 1995, p. 48), is the most studied aspect of group climate. In general, a higher level of cohesion is related to better member outcome (Budman et al., 1989; Kapp, Gleser, & Brissenden, 1964; Roether & Peters, 1972; Weiss, 1972; Yalom, Houts,
221
Zimerberg, & Rand, 1967). Studies of multidimensional aspects of group climate also show significant relationships with group member outcome. For example, MacKenzie, Dies, Coche, Rutan, and Stone (1987) found that successful groups, when compared with less successful groups, were more cohesive and engaged and were less defended, superficial, conflicted, and avoiding. Braaten (1989) showed that early ratings of higher engagement and lower avoidance were related to greater change in target complaints. Engaged and cohesive climates are associated with positive member outcome, whereas defended and conflicted climates are associated with negative member outcome. Tarrant and Kivlighan (2000) recently factor analyzed young adolescents' responses to the Group Climate Questionnaire (GCQ; MacKenzie, 1983) and found Active Engagement and Conflict-Distance factors of group climate. Researchers have historically examined group climate using one-time ratings of the climate (Kivlighan & Lilly, 1997). Group development theory (MacKenzie, 1994) and empirical evidence (Kivlighan & Lilly) suggest that the patterns of group climate change are better predictors of group member outcome that are one-time measures of group climate. For example, Kivlighan and Lilly found that a high-lowhigh pattern of engagement, a low-high-low pattern of conflict, and a cubic pattern of avoiding were related to group member outcome. Other research, however, suggests that simpler linear patterns of group climate development are good predictors of group member outcome (MacKenzie et al., 1987; Phipps & Zastowny, 1988; Sexton, 1996). Taken together, this research indicates that the pattern of climate development over time is a stronger predictor of group member outcome than is a one-time measure of climate. Therefore, we hypothesized that an increasing level of active engagement and a decreasing level of a conflict-distance would be positively related to group member outcome. Is Group Leadership Related to Group Climate? There are relatively few studies examining the relationship between group leadership and group climate. In an early study, Kratochvil and Vavrik (1976) found that group therapists' capacity for empathy, warmth, and sincerity was
222
KIVLIGHAN AND TARRANT
positively related to group cohesion. Hawks (1987), using sequential analysis, showed that a group leader facilitativeness and group (vs. individual) orientation led to a more productive group climate. Antonuccio, Davis, Lewinsohn, and Breckenridge (1987) found that depression treatment groups were more cohesive when the leader had an internal locus of control and was rated as warmer and less directive. These studies suggest that various aspects of group leadership are related to cohesion in therapeutic groups. Another group of studies examined the effect of group leader structuring on multiple aspects of group climate (Caple & Cox, 1989; Fuehrer & Keys, 1988; Kivlighan & Jauquet, 1990). These studies found that leader structure increased group cohesiveness and engagement and decreased defensiveness and conflict. Some early studies suggested that therapist structure should be sequenced, with initially high levels of structure being reduced as the group develops (e.g., Kinder & Kilmann, 1976). Recent studies suggest structure should be changed, not reduced, as groups develop (Kivlighan & Jauquet; Kivlighan, McGovern, & Corazzini, 1984). Summarizing across these studies, it appears that a positive leadership style and leader structuring are associated with more cohesive-engaged and less defensiveness-conflicted group climates. As noted above, group leader session intentions may provide a particularly effective way of examining the cognitive activity of the group leader. Building on the hypothesis about the effects group leadership and group climate have on group member outcome, we hypothesized that higher levels of group leader structuring and positive relationship would be associated with an increasing level of active engagement and a decreasing level of conflict-distance. Does Group Climate Mediate the Group Leadership-Group Member Outcome Relationship? In an early study, Bolman (1971) found partial support for a mediational role for group climate. Group members rated two aspects of group climate, withdrawal and tension, and also assessed group members' learning. Whereas congruence-empathy and openness were related to group tension, group tension was unre-
lated to either the member's own or other group members' learning. Group withdrawal, however, mediated the relationship between group leader perceptiveness and ratings of learning. Phipps and Zastowny (1988) also found partial support for the mediational role of group climate. A positive style of leadership was related to a more cohesive group climate, which in turn predicted better member outcome. Most studies of group leadership are studies of the individual group member's perception of the leader, because the member has been used as the unit of analysis. In the current study, the unit of analysis was the leader. Both Bolman's (1971) and Phipps and Zastowny's (1988) studies suffered from mono-method bias because group members rated the leader, group climate, and outcome. We attempted to improve on this design by having group leaders rate their intentions and having group members rate group climate and outcome. Summary of Research Questions and Hypotheses We initially identified the dimensions that describe group therapist session intentions. We also tested the following hypotheses depicted in Figure 1: (a) Group leaders' intentions to structure the group sessions and form positive relationships with group members will be related to increasing levels of group active engagement and decreasing levels of conflict-distance; (b) increasing levels of active engagement and decreasing levels of conflict-distance will be associated with positive group member outcome; and (c) there will be no significant relationships between group leaders' intentions to structure the group sessions and form positive relationships with group members and group member outcome when group climate is controlled. Method
Group Members Group members were 233 youths (103 boys and 130 girls) in the custody of the Division of Family Services (DFS), representing 32 counties in the State of Missouri, with a mean age of 14.15 years {SD = 1.07). Participant racialethnic breakdown was as follows: Caucasian 46% (n = 108), African American 34% (n =
MEDIATIONAL ROLE OF GROUP CLIMATE
223
Positive Relationship Increasing Active Engagement
Decreasing Conflict-Distance Structure Figure 1. Theoretical path model for the relationships among group leader session intentions, linear changes in group climate, and group member treatment satisfaction. N = 80.
80), Native American 8%, Hispanic 3%, and Asian American 1%. Group members reported living in a residential facility (43%), foster home (37%), at home (8%), or with a relative (11%). Group members had been in DFS custody an average of 3 years and 4 months and had changed placements an average of 2.09 times in the last year (SD = 1.70). Although the majority of group members reported never having been charged with a crime (n = 163; 70%), 67 group members (29%) indicated that they had been formally charged with a crime. The youths in these groups experienced a large number of problems. In the last 3 months, 48% of the youths reported using tobacco, 34% acknowledged consuming alcohol, 30% reported using marijuana, and 15% admitted using inhalants. A large percentage of the youths acknowledged being victimized, including 83% who acknowledged physical abuse. School problems were prevalent among the group members. Of the youths in this study, 86 indicated that they repeated a grade in school, and 159 reported being suspended or expelled from school (68%). Over half of the youths (53%) indicated that they had had consensual sex. The mean age of their first sexual experience was 12.07 years. Approximately one third of the youths in this study indicated that they were currently sexually active (n = 75; 32%).
Group Leaders Group leaders were 41 adults throughout the state of Missouri who were the only leader for a Choices group. Ninety percent of them had previously facilitated three or fewer Choices groups. Thirty-five group leaders were women (83%), and six were men (14%). Their ages ranged from 16 to 52 years, with a mean of 32.9 (SD = 8.3). Group leaders were predominately Caucasian (n = 36; 86%). African American group leaders represented 10% of the sample (n = 4). The majority of group leaders indicated that their highest educational level was a bachelor's degree (n = 22; 52%); eight of the leaders had completed some college (19%), four had a high school diploma (10%), four had an associates degree (10%), and one had a master's degree (2%).
Treatment Groups The Choices Independent Living Program is a statewide group intervention for youths ages 13-15 years who are in the custody of the DFS. Choices groups are 2 hr in length and span 8 weeks. Treatment manuals for the Choices groups were developed by Jeffrey M. Tarrant. Table 1 contains an outline of the Choices curriculum. Group sessions generally consist of
224
KIVLIGHAN AND TARRANT
Table 1 Session Outline for Choices Groups Session no. & title
Session content
Session 1: Getting Started Session 2: Values Clarification Session 3: Decision Making Session 4: Communication Skills Session 5: Anger Management Session 6: Drugs/Alcohol Session 7: Sexual Responsibility Session 8: Closing
Goals: Become familiar with the content and purpose of the groups as well as develop a familiarity and rapport with the group leaders and other group members. Goals: Identify and clarify personal values and examine the role that peer pressure plays in shaping values. Goals: Explore the importance and scope of decision making in daily life, describe decision-making difficulties, and relate decision making to values. Goals: Identify and practice good listening skills. Goals: Explore personal anger responses and increased awareness of physical, cognitive, and emotional reactions to anger. Goals: Disseminate basic facts about substance use, estimate risk of substance use, identify pros and cons of substance use. Goals: Discuss different types of relationships, distinguish between sex and intimacy, learn ways to express love and sexual feelings without engaging in sexual intercourse. Goals: Reflect on group experience; identify learning.
structured group exercises, didactic instruction, and group discussion and interaction around the session topic. Although the sessions are semistructured, group leaders are trained to emphasize member-to-member interaction. Measures Youth Client Satisfaction Questionnaire (YCSQ). The YCSQ (Shapiro, Welker, & Jacobson, 1997) was designed as a child-centered approach to consumer satisfaction for use with youths 11 years and older. The YCSQ is composed of 14 items scored on a 4-point Likert scale ranging from 0 (not at all) to 3 (a great deal). All items included in this scale met three criteria: (a) test-retest reliability correlation equal to or greater than .50, (b) part-whole correlation greater than or equal to .40, and (c) correlation with a validation item greater than or equal to .30 (Shapiro et al.). Average correlations for the retained items were .67 for testretest reliability, .65 for part-whole correlation, and .48 for validity. Total scores had a testretest reliability coefficient of .92. Factor analysis of the YCSQ revealed two independent factors: Relationship With Therapist and Benefits of Therapy. Both factors had a Cronbach's alpha of .85. Validity for the YCSQ was explored through correlations with client demographic and treatment variables. Boys and younger clients reported more satisfaction with treatment than girls or older
youths. YCSQ ratings were significantly correlated with parents' satisfaction with the youths' treatment, parents' ratings of the youths' benefit from treatment, and pre- to posttreatment changes on the Child Behavior Checklist and Global Assessment of Functioning. In a larger evaluation of the Choices groups, the Benefits of Therapy factor was the single best indicator of group member outcome. The Benefits of Therapy factor was correlated with pre- to postgroup decreases in the following: (a) expected involvement in illicit drug use, aggressive and illegal behavior, risky sexual activities, heavy drinking, high-risk sports, and risky academic and work behaviors, as measured by the Cognitive Appraisal of Risky Events Scale (Fromme, Katz, & Rivet, 1997); (b) assault, indirect hostility, irritability, negativism, resentment, suspicion, and verbal hostility, as measured by the Buss Durkee Hostility Inventory (Buss & Durkee, 1957); and (c) intrapersonal distress, somatic distress, interpersonal relations, social problems, and behavioral dysfunction, as measured by the Youth Outcome Questionnaire (Wells, Burlingame, Lambert, & Hoag, 1996). In this same evaluation study, the Relationship With Therapist factor was significantly correlated with session attendance. Group Climate Questionnaire—Short Form (GCQ-S). Group climate, defined as a participant's perception of the group atmosphere, was measured with the group participants' ratings on the short form of the GCQ (MacKenzie,
MEDIATIONAL ROLE OF GROUP CLIMATE
1983). The GCQ-S includes 12 items reported on a 6-point Likert scale indicating degree of agreement ranging from 1 (not at all) to 6 (extremely). Factor analysis of these items resulted in a three-factor structure: Engagement (degree of cohesion and work orientation in the group), Avoidance (the degree to which individuals rely on the other group members or leaders), and Conflict (interpersonal conflict and distrust). Interscale correlations were as follows: Avoidance and Engagement, -.44; Conflict and Engagement, -.18; and Conflict and Avoidance, .30 (MacKenzie, 1983). Example GCQ-S items include "The members tried to understand why they do the things they do, tried to reason it out" (Engagement); "The members avoided looking at important issues going on between themselves" (Avoidance); and "There was friction and anger between the members" (Conflict). The GCQ-S has been used in studies to assess climate differences across groups (Kanas & Barr, 1986; MacKenzie et al., 1987) and to examine the relationship between group member interpersonal problems and perceptions of group climate (Kivlighan & Angelone, 1992). Tarrant and Kivlighan (2000) found a twofactor structure for the GCQ-S using ratings from adolescents in the Choices groups. The two factors were labeled Active Engagement (AE) and Conflict-Distance (CD). The AE factor accounted for 26% of the total variance, whereas the CD factor accounted for an additional 24%. Alpha reliabilities for the two factors were good, .92 and .90 for the AE and CD factors, respectively. Validity for the AE and CD factors was established by examining how these dimensions changed across time and by examining the relationship between group members' perception of the group leader and their perception of AE and CD. As hypothesized, AE increased and CD decreased over time. Also as hypothesized, AE was positively related to leader engagement and negatively related to leader disengagement. CD showed the opposite pattern of relationships with leader engagement and disengagement. These findings supported the construct validity of the AE and CD factors. TSI (Stiles et al, 1996). The TSI consists of 19 items scored on a 5-point Likert scale ranging from 1 (not at all) to 5 (very much). The TSI is completed at the end of each group
225
session by the group facilitator. The TSI offered a way to measure how much each of the 19 intentions was focused on in the group session. Intentions include items such as the following: set limits, get information, and examine the relationship. Factor analysis of the TSI by Stiles et al. (1996) resulted in eight factors that represented distinct therapeutic foci. The eight factors were as follows: Treatment Context Focus, Session Structure Focus, Affect Focus, Obstacles Focus, Encouraging Change Focus, Behavior Focus, Cognition-Insight Focus, and Mixed/ Complex Item. A number of findings from Stiles et al.'s study support the validity of the TSI factors. First, TSI factors distinguished between psychodynamic-interpersonal and cognitive-behavioral treatments. Second, TSI factors showed theoretically consistent patterns of change across time. Third, the TSI factors showed the expected relationships with observer ratings of diverse treatments. Because the TSI was never used in group treatment, we conducted a factor analysis to examine structure of items in this specific context.
Procedure Two 8-hr training sessions were held for Choices supervisors and group facilitators. This training was mandatory and was composed of three parts. First, group facilitators were shown the curriculum. Part 2 of the training consisted of in-depth explanations of the research portion of this project. Issues discussed included the following: introduction of the research to group members, confidentiality-anonymity, data collection, as well as summaries and justifications for all measures included in the study. The third part of this training consisted of group leader skills training. This portion of the training focused on their role as a group facilitator and provided information on content versus process in structured groups as well as the importance of consistency and professionalism. Group leaders were given copies of the group leader code of ethics (ASGW Ethics Committee, 1998). This training also addressed issues pertinent to structuring of the first group session, including the importance of arriving early, seating arrangement issues, options for group introductions, and ways to discuss group rules with their group. Training also included information on how to deal with problem group members. Last,
226
KIVLIGHAN AND TARRANT
group leaders were given a self-assessment to complete to identify growth areas as a group facilitator. Group leaders and group members filled out consent forms before the meeting of the first group session. Leaders and members were informed that the data were anonymous and could not be linked to a specific person. They were also informed that the data would be used to examine overall program effectiveness and would not focus on any particular group leader or group member. All groups met for 2 hr a week for 8 weeks. Group members completed the GCQ-S at the end of each session they attended. At termination of counseling, all participants were asked to fill out the YCSQ. Group leaders filled out the TSI after each group session. Because we were interested in climate as a group phenomenon, we aggregated the GCQ-S scores for the individual group members for each session by averaging (calculating the mean) across all group members attending the session for both the AE and CD factors. We calculated a reliability index to determine if group climate scores were consistent across the group members for a group session. A coefficient alpha for AE and CD was calculated by specifying individual group members as items and group sessions as occasions. Internal consistencies for the AE and CD factors were .89 and .87, respectively, indicating that the group members had a consistent view of the sessions' climate. We also wanted to calculate overall measures of success for each group studied. Individual group members' scores on the two YCSQ factors were aggregated by averaging (calculating the mean) across all members of a Choices group. Therefore, each group had two scores: the average level of Relationship With Therapist and Benefits From Treatment. Reliabilities for the Relationship With Therapist and Benefits From Treatment scores were calculated as indicated above for group climate. Internal consistencies for the group level of relationship with therapist and benefits from treatment were .83 and .85, respectively, indicating that the group members had a consistent view of their satisfaction with the group experience.
Results Group leader ratings for the 19 TSI items from 212 group sessions were subjected to an exploratory analysis (iterative principal-components extraction). Because the TSI is intended for use as a session-by-session measure, the session was the unit of analysis. According to Elliott and Wexler (1994), "exploratory factor analysis is a descriptive technique rather than an inferential statistical method; therefore, nonindependence of sessions within cases is not critical" (p. 168). The Kaiser-Meyer-Olkin measure of sampling adequacy (Kaiser, 1974) was .87, which indicated factor analysis was appropriate. Four factors met the Kaiser-Guttman retention criterion of eigenvalues greater than 1.0. Cattell's scree test also suggested a four-factor solution. These four factors explained 63% of the total item variance. We used an orthogonal rotation because the oblique rotation indicated only small correlations among the four factors. Table 2 contains factors' loading for the items and means and standard deviations for the 19 TSI items. Two criteria were used to determine an item's placement on a factor: (a) the item had to load .50 or better on the targeted factor and (b) an item's loading on a targeted factor had to be at least .20 greater than its loading on any of the other three factors. Using these criteria, seven TSI items loaded on the first factor. The first factor was named Therapeutic Work Focus because the feeling-awareness, insight, challenge, and change items had the highest loadings on this factor. These intentions are associated with working toward therapeutic change. After rotation, the first factor accounted for 25% of the item variance. The second factor, which accounted for 14% of the variance, was composed of three TSI items. This factor was named Group Structure Focus because the focus, behaviors, and clarify intentions suggest that the group leader was intending to help the group members develop appropriate group norms. The third factor consisted of three TSI items and accounted for 13% of the total item variance. The intentions of therapist needs, relationship, and resistance seemed to capture the interpersonal focus of the groups and were therefore named Interpersonal Focus. The fourth factor included two items and accounted for 11% of
227
MEDIATIONAL ROLE OF GROUP CLIMATE
Table 2 Rotated Pattern Matrix, Item Means, and Standard Deviations for the Therapist Session Intention Form (TSI) Item 12. 13. 17. 14. 15. 11. 3. 7. 5. 10. 6. 19. 18. 16. 9. 4. 1.
Feelings-awareness Insight Challenge Change Reinforce change Self-control Give information Hope Focus Behaviors Clarify Therapist needs Relationship Resistance Cognitions Support Set limits
Safe Therapeutic Structuring work group Interpersonal atmosphere .80 .77 .74 .69 .62 .60 .58 .56 .12 .21 .25 -.02 .33 .50 .40 .27 -.06
-.01 .14 .23 .17 .42 .36 .14 .06 .82 .73 .63 .03 .26 .29 .45 .14 .37
.13 .22 .30 .18 .18 .08 -.11 .36 -.04 .31 .15 .82 .58 .57 .52 -.08 .22
.09 .04 -.12 .27 .23 .25 .13 .43 .35 -.05 .25 -.01 .27 .21 -.00 .77 .68
M
SD
3.65 3.65 3.42 3.72 3.65 3.96 4.13 3.77 4.14 3.70 3.96 .78 3.47 3.07 3.15 4.30 3.92
1.13 1.11 1.24 1.19 1.18 0.93 0.91 1.16 0.97 1.20 1.05 1.26 1.14 1.29 1.24 0.73 1.02
Note. N = 212 group sessions. Items 2 (get information) and 8 (cathart) were complex items not loading cleanly on any factor. Boldface type indicates which items composed the derived scales. TSI items were scored on a 5-point Likert scale ranging from 1 (not at all) to 5 {very much).
the item variance. The combination of support and set limits intentions suggests that the group leaders were attempting to limit acting out behavior while also being supportive to group members. Therefore, we labeled this factor Safe Atmosphere Focus. Cronbach's coefficient alphas were calculated to examine the internal consistency of the Therapeutic Work Focus, Group Structure Focus, Interpersonal Focus, and Safe Atmosphere Focus factors. Acceptable coefficient alphas were obtained for all factors, supporting their internal consistency. The reliability estimates were .87, .75, .68, and .56 for the Therapeutic Work Focus, Group Structure Focus, Interpersonal Focus, and Safe Atmosphere Focus factors, respectively. We used hierarchical linear modeling (HLM) to estimate linear growth coefficients for the AE and CD factors from the GCQ-S for each Choices group. Growth modeling, with HLM, offers a powerful method for analyzing repeated measures group process data. In this study, a linear growth model was fit to the aggregated AE and CD data for each Choices group. The linear parameters (slopes) from these initial
models were used as the group climate data in a path analysis to test the hypotheses. Table 3 displays the means and standard deviations for standardized linear parameter (slope) coefficients for the AE and CD factors as well the means and standard deviations for the TSI factors and the two YCSQ factors. AE increased 0.1 scale points per session and CD decreased 0.07 scale points per session. Across the eight sessions, AE increased about 1 scale point, from a rating of approximately 4 to a rating of approximately 5 on the 6-point Likert scale. Across the eight sessions, CD increased about 0.5 scale points, from a rating of approximately 2 to a rating of approximately 1.5 on the 6-point Likert scale. With their community mental health center sample, Shapiro et al. (1997) found ratings on the Relationship With Therapist factor and the Benefits From Treatment factor averaged 2.34 and 2.08, respectively. The Choices group members had comparable ratings on the Relationship With Therapist factor but relatively lower scores on the Benefits From Treatment factor. The Relationship With Therapist factor mean was almost one point above the scale
228
KIVLIGHAN AND TARRANT
Table 3 Means and Standard Deviations for and Correlations Among Therapist Session Intentions, Group Climate, and Treatment Outcome Measure 1. 2. 3. 4. 5. 6. 7. 8.
TSI: Therapeutic Work Focus TSI: Group Structure Focus TSI: Interpersonal Focus TSI: Safe Atmosphere Focus GCQ: Active Engagement GCQ: Conflict-Distance YCSQ: Relationship With Therapist YCSQ: Benefits of Therapy
M 2.65 2.06 3.47 3.65 0.10 -0.07 2.37 1.50
SD
1
2
3
4
5
6
7
8
1.26 1.42 1.66 0.96 0.51 0.90 0.35 0.53
— .22 .25 .22 .50 .00 .42 .42
— .19 .15 .15 .46 .35 .02
— .16 .21 .31 .35 .06
— .64 .13 .45 .45
— .04 .18 .67
.35 .05
.27
—
Note. N = 43 groups. TSI = therapist session intentions—scores ranged from 1 (not at all) to 5 (very much); GCQ = Group Climate Questionnaire—Active Engagement and Conflict-Distance scores are slope coefficients derived from a 6-point Likert scale, so Active Engagement increases .1 scale points per session and Conflict-Distance decreases .07 scale points per session; YCSQ = Youth Client Satisfaction Questionnaire—Relationship With Therapist and Benefits of Therapy scores ranged from 0 (not at all) to 3 (a great deal).
midpoint (1.5), whereas the average Benefits From Treatment factor score was half a point above the scale midpoint. This suggests that the Choices group members liked their group leaders and benefited from treatment. A within-subjects repeated measures multivariate analysis of variance revealed that there was a significant difference in the relative frequency of intention use, F(3, 110) = 61.76, p < .001; TJ2 = .63. Pairwise comparisons showed that the Choices group leaders had a higher focus on interpersonal and safe atmosphere intentions than on therapeutic work intentions. In addition, therapeutic work intentions received a higher focus than did group structure intentions. We used causal modeling procedures to examine the relationship of group leader intentions and group climate with group member outcome. We used path analysis in this study rather than structural equation modeling because of our use of a single measure for each construct and because of our sample size. According to Asher (1984), path analysis makes use of bivariate and partial regression techniques to calculate path coefficients. Bivariate correlations are used when a path variable is causally related to only one other variable in the model. The path coefficient (partial r) indicates the strength of a given path coefficient, controlling for the variance that results from other causal variables (Mallinckrodt, 1996). The proposed path model consisted of three blocks of causally ordered predictor variables. The path model was recursive; that is, the direction of effects was one-way, from left to right. The
initial block consisted of the four foci of therapist session intentions that were exogenous to the model. The second block consisted of the group climate dimensions. The final block of the model contained the two factors of treatment satisfaction. The hypothetical model presented in Figure 1 identifies two types of leader intentions. The factor analysis of the TSI, however, revealed four factors of therapist intentions. The Group Structure factor appeared to correspond to the hypothesized structure dimension, whereas the Safe Atmosphere factor appeared to correspond to the hypothesized positive relationship dimension. In addition, in this sample the Benefits of Therapy and Relationship With Therapist factors from the YCSQ represented independent aspects of client outcome. Therefore, the path model predicted these two aspects of group member outcome. We initially ran a complete path model to examine all possible relationships between therapist intentions, group climate, and group member outcome. The significant paths from this complete model were retained in our final path model. These significant paths are depicted in Figure 2. The model depicted in Figure 2 provides a good fit to the data, as indicated by the chi-square goodness-of-fit index, ;^(20, N = 43) = 26.86, p = .22); the Bentler-Bonnett Nonnormed Fit index (.90); and the Comparative Fit index (.92). In the path model, therapeutic work, safe atmosphere, group structure, and interpersonal intentions were independent (no significant covariance) aspects of the group
MEDIATIONAL ROLE OF GROUP CLIMATE
229
^•0.87
E5 Figure 2, Path model for the relationships among the four dimensions of group leader session intentions, linear changes in two dimensions of group climate, and two dimensions of group member treatment satisfaction. N = 43. E5, E6, E7, and E8 refer to errors associated with respective variables. * indicates a significant path at the p < .05 level.
leaders' cognitive planning activities. Likewise, AE and CD were independent aspects of the groups' climate, whereas benefits of therapy and relationship with therapist were independent aspects of group outcome. In partial support of the hypotheses, the safe atmosphere intentions (positive therapeutic relationship) were significantly related to a group climate increasing in AE—j3 = .40; r(41) = 3.60, p < .05— and group structure intentions (therapist structuring) were significantly related to a climate decreasing in conflict and distance—/3 = .42; t(41) = 3.15, p < .05. In addition, therapeutic work intentions were significantly related to a group climate increasing in active engagement— /3 = .58; f(41) = 5.26, p < .05—and interpersonal intentions were marginally related to a climate decreasing in conflict and distance— j3 = .24; f(41) = 1.81, p = .08. The hypothesized relationships between group climate and group member outcome were supported. To be specific, an increase in AE across sessions was significantly related to the members in a group perceiving more benefits of therapy—/3 = .65; /(41) = 5.56, p < .05. A decrease in CD across sessions was significantly related to the members in a group being more
satisfied with their relationship with therapist— j3 = .58; K41) = 2.52, p < .05. The hypothesized mediational role for group climate was mostly supported. An increase in AE completely mediated the relationship between the therapeutic work and safe atmosphere intentions and members in a group perceiving more benefits of therapy. To be specific, when increases in AE were controlled, there were no significant direct effects for the therapeutic work and safe atmosphere intentions on members in a group perceiving more benefits of therapy. Likewise, decrease in CD completely mediated the relationship between the group structure and interpersonal intentions and members in a group perceiving a better relationship with therapist. When controlling for decreases in CD, there were no significant direct effects for the group structure and interpersonal intentions on members in a group perceiving a better relationship with therapist. The therapeutic work—^ = .36; t(4l) = 2.88, p < .05—and safe atmosphere—/3 = .34; r(41) = 2.52, p < .05—intentions, however, did have a significant direct relationship with members in a group perceiving a better relationship with therapist.
230
KIVUGHAN AND TARRANT
Discussion The results of this study support many of Yalom's (1995) hypotheses about group leadership and group climate and provide direction for how a group leader builds a therapeutic group climate and how this climate affects group member outcome. The results also suggest that group leader intentions are an important foci for understanding group leadership. The principalfactors analysis of the TSI revealed four dimensions of group leader intention compared with eight dimensions found in individual therapy (Stiles et al., 1996). The four dimensions provide a parsimonious and relatively complete description of intentions in group therapy. Despite the differences in the number of dimensions found, there were some striking similarities between the TSI dimensions in group and individual therapy. The focus and clarify intentions formed a session structure dimension in Stiles et al.'s (1996) study; in this study, the group structure dimension had these two intentions plus the behaviors intention. The behaviors intention involves providing group members with feedback about their inappropriate behavior. The addition of the behaviors intention to the focus and clarify intentions suggests that structuring is a broader activity in group treatment. Therapist needs, relationship, resistance, and challenge intentions denned an obstacles focus in Stiles et al.'s (1996) study. Without the challenge intention, the remaining intentions seem to represent an interpersonal focus (Yalom, 1995). A major task of the group therapist is to encourage the group members to focus on the here-and-now relationship. In individual therapy, the relationship may become a focus only when problems develop. This might explain why challenge is coupled with the interpersonally oriented intentions in individual therapy. The therapeutic work focus seemed to combine the affect, encouraging change, behavior, and cognition-insight foci found in Stiles et al. (1996). Individual therapists seem to make more narrow and specific use of these intentions, whereas group leaders use these intentions in a more global and undifferentiated manner. In group treatment, interpersonal intentions are predominant, with therapeutic work intentions playing a secondary, less differentiated role.
In Stiles et al.'s (1996) study, the set limits intention was a complex item with loadings on several dimensions. The set limits intention loaded on the treatment context focus dimension with the get and give information intentions. In this study, the support and set limits intentions formed a safe environment focus. This safe environment focus had the highest mean rating of the four foci. It is not surprising that providing a safe environment is a predominant theme for group leaders working with troubled youths. For these youths to feel safe, they must have clear rules and expectations in the context of a warm, supportive, and empathic environment. In the zero-order correlations, all four foci were related to relationship with therapist, and the therapeutic work and safe environment foci were related to benefits of therapy. This finding replicates group leadership research reviewed by Dies (1994), which highlighted the important role structure plays in group member outcome. A group structure focus was related to a positive relationship with the therapist but not with perceptions of benefit from the group. It is important to remember that in these groups a positive relationship with the therapist was related to better group attendance. Members can obviously benefit only from groups that they attend. Members feel more positive about group leaders who structure the discussion, solicit elaboration of material, and give feedback about group member behaviors. Dies (1994) also said that it was clear that members benefited from a positive leadership approach. The relationship between the safe environment and the benefit members receive supports Dies's conclusion about a positive leadership style. Focusing on therapeutic work was negatively related to both benefit from the group and relationship with the group leader. The therapeutic work intentions involve a focus on an individual group member. This suggests that members are less satisfied with the group and benefit less from the group when the leader focuses on individual group members. Recent research suggested that dynamic, as opposed to static (one-time), measures of group climate provide better predictions of group outcome (Kivlighan, Coleman, & Anderson, 2000; Kivlighan & Lilly, 1997). A group climate that was increasingly active and engaged was related to members seeing the group as more beneficial.
MEDIATIONAL ROLE OF GROUP CLIMATE In addition, a group climate that decreased in
conflict and distance was related to members having a more positive relationship with the leader. Because rated benefit was related to decreased anger, decreased risky behavior, and better general functioning and the relationship with the leader was related to better attendance, these group climate findings take on added weight. It is interesting that an increasingly active and engaged group climate uniquely predicted member-rated benefit, whereas a climate decreasing in conflict and distance uniquely predicted members' feelings about the leader. Group interactions (Bales, 1950) and group member roles (Forsyth, 1999) can be classified as task or socioemotionally oriented. Task interactions involve the activities that must be accomplished by the group members, and socioemotional interactions involve interpersonal relations within the group. MacKenzie (1983, p. 159) argued that group climates describe "features of the therapeutic environment that encourage compatible types of interpersonal events." We suggest that an increasingly engaged and active group climate encourages task-oriented roles and interactions (e.g., learning and practicing communication skills and anger management techniques). Member-rated benefit and the associated problem reduction are related to engaging in these tasks. On the other hand, a climate decreasing in conflict and distance encourages socioemotional interactions and roles. Socioemotional interactions and roles "sustain the emotional bonds linking members to one another and the group" (Forsyth, p. 8). A better relationship with the leader and better group attendance are associated with the socioemotional interactions that occur in a decreasing climate of conflict and distance. Because group climate has a strong relationship with group member outcome, it is important to know what the leader can do to influence group climate. The results of this study begin to clarify how aspects of group leadership are related to dimensions of group climate. A therapeutic work focus was negatively related and a safe environment focus was positively related to an increasingly active and engaged group climate. A group structure focus was significantly related and an interpersonal focus was marginally related to a group climate decreasing in conflict and hostility.
231
The intentions involved in therapeutic work consist mainly of individually oriented change techniques. As in Hawks (1987), an individual orientation is associated with a less productive group climate. Focusing on a safe environment, however, is associated with members increasing their active engagement in the group. The safe environment focus is best described as setting therapeutic norms in the context of a warm and supportive relationship. Both Kratochvil and Vavrik (1976) and Antonuccio et al. (1987) showed the importance of group leader warmth in creating positive group environments. In addition, Kivlighan and Jauquet (1990) found that high-quality goal setting, an important aspect of norm development, was related to an engaged group climate. To summarize, group members will increase their active involvement with the group when group leaders refrain from doing individual therapy in the group and actively set goals and norms while maintaining a warm and supportive environment. The importance of group structure in decreasing conflict and distance can be explained by Bednar, Melnick, and Kaul's (1974) risk-responsibility model. The model suggests that structure reduces the anxiety and risk involved in participating in the group, letting group members engage in less anxiety-related behaviors (e.g., avoidance). The relationship between the group structure intentions supports the riskresponsibility model. The study's most important finding is the support for the mediational role of group climate in the relationship between group leader intentions and group member outcome. As Yalom (1995) suggested, the group leader's major task is to create a therapeutic group climate. The therapeutic climate will foster interactions that can result in positive group member outcome. Unlike individual treatment, where the relationship between the client and therapist is tantamount, in group treatment leaders should probably de-emphasize their relationships with the individual group members and focus on creating a therapeutic group climate. The results of this study confirm and expand on the findings in Bolman (1971) and Phipps and Zastowny (1988). The strong mediational role for group climate found in this study is surprising given the semistructured nature of the groups examined. Yalom's (1995) writings about group climate
232
KIVLIGHAN AND TARRANT
are in the context of unstructured, open-ended, interactional type of groups. One would presume that the group climate should be less important in more structured groups because the activities should carry more of the therapeutic weight. However, group climate was an important mediator even in these semistructured groups. This finding suggests that Yalom's proposition about the mediational role of group climate is robust and not simply limited to unstructured therapy type groups. The evidence for the mediational role of group climate was much stronger in this study than in either Bolman's (1971) or Phipps and Zastowny's (1988) study. There are several reasons why the results of this study are more robust than those of Bolman and Phipps and Zastowny. First, group climate was examined more broadly in this study by including both positive (active and engaged) and negative (conflict and distance) aspects of group climate and examining changes in group climate over time. Second, group member outcome was also examined broadly in this study by including both task-related (member perception of benefit) and socioemotional-related (relationship with therapist) aspects of outcome. Third, the aggregation of AE and CD scores across group members in this study provided a measure of climate at the group level. Fourth, measuring group leaders' intentions provides a better characterization of group leader activity than does measuring group members' perceptions of the leader's behaviors or attitudes. There are a number of limitations of this study that are important to mention in that they may qualify our findings. First, the type of groups examined (semistructured) differed from that on which Yalom's (1995) model was based (unstructured). As noted above, however, we believe that our findings speak to the robustness of Yalom's model. Even in semistructured groups, creating a therapeutic group climate seems to be the priority activity for the group leader. Second, the group leaders were paraprofessionals with a limited educational background, little formal group training, and little group leadership experience. It is not clear to what extent these findings can be generalized to professionally led groups. It is important to note that these groups were not created for this study. Rather, we examined a rather large state program that was using a group treatment format.
We believe that group therapy professionals and group researchers can have a beneficial impact on these type of real-world groups both through leader training and evaluation research. Third, although 43 leaders represents a relatively large number of groups when compared with published group treatment studies, on a statistical level the number is rather modest. Fourth, because group members rated both group climate and their own outcome, monomethod bias exists. Mono-method bias, however, cannot explain the relationships between group leader intentions and group climate or between group leader intentions and group member outcome. Future research could use observer ratings of group climate and group member outcome to guard against monomethod bias. Fifth, we used linear models to examine dynamic changes in group climate. More complex models of change in group climate are statistically and theoretically possible. Future research could compare the predictive power of different models of group climate change. Sixth, several questions can be raised about the construct validity of the measures used in this research. The GCQ has been widely used in group therapy research but has seldom been used with a young adolescent population in group treatment. Whereas Tarrant and Kivlighan (2000) found a two-factor structure for the GCQ with this young adolescent population, this finding needs replication. The YCSQ is a relatively new measure, but it does have good initial reliability and validity information. On the positive side, the YCSQ was validated with a young adolescent age group, a large percentage of whom were receiving group treatment. Although the development of the TSI was rigorous and thorough, this instrument was never used before by group leaders. The factor structure of the TSI made theoretical and practical sense, and the correlations between the TSI dimensions and both group climate and group member outcome serve to support the validity of the TSI with this population. In addition, the TSI is a self-report measure of group leader intentions; it is not clear how these intentions relate to in-session group leader behavior. Future research could provide more information about the validity of the TSI in a group counseling session. Despite these limitations, the results of this study were quite encouraging with regard to
MEDIATIONAL ROLE OF GROUP CLIMATE
Yalom's (1995) writings about the role of the group leader. As suggested by Yalom, the leader's primary task is building and maintaining a therapeutic group climate. Future research could help refine Yalom's model by examining aspects of group leadership other than intentions, exploring more complex models of group climate change, and addressing perspectives on client outcome other than client self-report.
References Antonuccio, D. O., Davis, C , Lewinsohn, P. M., & Breckenridge, J. S. (1987). Therapist variables related to cohesiveness in a group treatment for depression. Small Group Behavior, 18, 557-564. ASGW Ethics Committee. (1998). Association for Specialists in Group Work best practice guidelines. Journal for Specialists in Group Work, 23, 237-244. Asher, H. B. (1984). Causal modeling (2nd ed.). Newbury Park, CA: Sage. Bales, R. F. (1950). Interaction process analysis: A method for the study of small groups. Reading, MA: Addison-Wesley. Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173-1182. Bednar, R., Melnick, J., & Kaul, T. (1974). Risk, responsibility and structure: A conceptual framework for initiating group counseling and psychotherapy. Journal of Counseling Psychology, 21, 31-37. Bolman, L. (1971). Some effects of trainers on their T groups. The Journal of Applied Behavioral Science, 7, 309-325. Braaten, L. J. (1989). Predicting positive goal attainment and symptom reduction from early group climate decisions. International Journal of Group Psychotherapy, 39, 377-387. Budman, S. H., Soldz, S., Demby, A., Feldstein, M., Springer, T., & Davis, M. S. (1989). Cohesion, alliance and outcome in group psychotherapy. Psychiatry, 52, 339-350. Buss, A. H., & Durkee, A. (1957). An inventory for assessing different kinds of hostility. Journal of Consulting Psychology, 21, 343-349. Caple, R. B., & Cox, P. L. (1989). Relationships among group structure, member expectations, attraction to the group, and satisfaction with the group experience. Journal for Specialists in Group Work, 14, 16-24. Coyne, R. K., Harvill, R. L., Morganett, R. S., Morran, D. K., & Hulse-Killacky, D. (1990). Effective
233
group leadership: Continuing the search for greater clarity and understanding. Journal for Specialists in Group Work, 15, 30-36. Dies, R. R. (1994). Therapist variables in group psychotherapy research. In A. Fuhriman & G. M. Burlingame (Eds.), Handbook of group psychotherapy (pp. 114-154). New York: Wiley. Elliott, R., & Wexler, M. M. (1994). Measuring the impact of sessions in process-experiential therapy of depression: The Session Impact Scale. Journal of Counseling Psychology, 41, 166-174. Forsyth, D. R. (1999). Group dynamics (3rd ed.). Belmont, CA: Wadsworth. Fromme, K., Katz, E. C , & Rivet, K. (1997). Outcome expectancies and risk-taking behavior. Cognitive Therapy and Research, 21, 421-442. Fuehrer, A., & Keys, C. (1988). Group development in self-help groups for college students. Small Group Behavior, 19, 325-341. Hawks, I. K. (1987). Facilitativeness in small groups: A process-oriented study using lag sequential analysis. Psychological Reports, 61, 955-962. Hill, C. E., & O'Grady, K. E. (1985). List of therapist intentions illustrated in a case study and with therapists of varying theoretical orientations. Journal of Counseling Psychology, 32, 3-22. Hill, C. E., Helms, J. E., Tichenor, V., Speigel, S. B., O'Grady K. E., & Perry, E. P. (1988). Effects of therapist response modes in brief psychotherapy. Journal of Counseling Psychology, 35, 222-233. Hines, P. L., Stockton, R., & Morran, D. K. (1995). Self talk of group therapists. Journal of Counseling Psychology, 42, 242-248. Kaiser, H. F. (1974). An index of factorial simplicity. Psychometrika, 39, 31-36. Kanas, N., & Barr, M. A. (1986). Process and content in a short-term inpatient schizophrenic group. Small Group Behavior, 17, 355-363. Kapp, F. T., Gleser, G., & Brissenden, A. (1964). Group participation and self-perceived personality change. Journal of Nervous and Mental Disease, 139, 255-265. Kinder, B. N., & Kilmann, P. R. (1976). The impact of different shifts in leader structure on the outcome of internal and external group participants. Journal of Clinical Psychology, 32, 857-863. Kivlighan, D. M., Jr., & Angelone, E. O. (1992). Interpersonal problems: Variables influencing participants' experience of group therapy. Journal of Counseling Psychology, 39, 468-472. Kivlighan, D. M., Jr., Coleman, M. N., & Anderson, D. C. (2000). Process, outcome and methodology in group counseling research. In S. D. Brown & R. W. Lent (Eds.), Handbook of counseling psychology (3rd ed., pp. 767-796). New York: Wiley. Kivlighan, D. M., Jr., & Jauquet, C. A. (1990). Quality of group member agendas and group session climate. Small Group Research, 21, 205-219.
234
KIVLIGHAN AND TARRANT
Kivlighan, D. M., Jr., & Lilly, R. L. (1997). Developmental changes in group climate as they relate to therapeutic gain. Group Dynamics: Theory, Research, and Practice, 3, 208-221. Kivlighan, D. M., Jr., McGovern, T. M., & Corazzini, J. G. (1984). The effects of the content and timing of structuring interventions on group therapy process and outcome. Journal of Counseling Psychology, 31, 363-370. Kivlighan, D. M., Jr., & Quigley, S. T. (1991). Dimensions used by experienced and novice group therapist to conceptualize group process. Journal of Counseling Psychology, 38, 415-423. Kratochvil, S., & Vavrik, P. (1976). Effect of therapist's qualities on cohesion and tension in group session. Ceskoslovenska Psychiatrie, 72, 332-336. MacKenzie, K. R. (1983). The clinical application of a group climate measure. In R. R. Dies & K. R. MacKenzie (Eds.), Advances in group psychotherapy: Integrating research and practice (pp. 159-170). Madison, CT: International Universities Press. MacKenzie, K. R. (1994). Group development. In A. Fuhriman & G. M. Burlingame (Eds.), Handbook of group psychotherapy (pp. 223-268). New York: Wiley. MacKenzie, K. R., Dies, R., Coche, E., Rutan, J. S., & Stone, W. N. (1987). An analysis of AGPA Institute groups. International Journal of Group Psychotherapy, 37, 55-74. Mallinckrodt, B. (1996). Change in working alliance, social support, and psychological symptoms in brief therapy. Journal of Counseling Psychology, 43, 448-455. Phipps, L. B., & Zastowny, T. R. (1988). Leadership behavior, group climate and outcome in group psychotherapy: A study of outpatient psychotherapy groups. Group, 12, 157-171. Richardsen, A. M., & Piper, W. E. (1986). Leader style, leader consistency, and participant personality effects on learning in small groups. Human Relations, 39, 817-836.
Roether, H. A., & Peters, J. J. (1972). Cohesiveness and hostility in group psychotherapy. American Journal of Psychiatry, 128, 1014-1017. Sexton, H. C. (1996). Exploring a psychotherapeutic change sequence: Relating process to intersessional and posttreatment outcome. Journal of Consulting and Clinical Psychology, 61, 128-136. Shapiro, J. P., Welker, C. J., & Jacobson, B. J. (1997). The Youth Client Satisfaction Questionnaire: Development, construct validation, and factor structure. Journal of Child Clinical Psychology, 26, 87-98. Stiles, W. B., Startup, M., Hardy, G. E., Barkham, M., Rees, A., Shapiro, D. A., & Reynolds, S. (1996). Therapist session intentions in cognitivebehavioral and psychodynamic-interpersonal psychotherapy. Journal of Counseling Psychology, 43, 402-414. Tarrant, J. M., & Kivlighan, D. M., Jr. (2000). An evaluation of the Choices group program. Unpublished manuscript, University of Missouri—Columbia. Weiss, B. J. (1972). Development of cohesiveness in marathon growth groups (Doctoral dissertation, University of Maryland, 1972). Dissertation Abstracts International, 32(10-B) 6065. Wells, M. G., Burlingame, G. M., Lambert, M. J., & Hoag, M. J. (1996). Conceptualization and measurement of patient change during psychotherapy: Development of the Outcome Questionnaire and Youth Outcome Questionnaire. Psychotherapy, 33, 275-283. Yalom, I. D. (1995). The theory and practice of group psychotherapy (4th ed.). New York: Basic Books. Yalom, I. D., Houts, P. S., Zimerberg, S. M., & Rand, L. (1967). Prediction of improvement in group psychotherapy. Archives of General Psychia-
try, 17, 159-168.