The purpose of Study 2 was to examine whether different types of student clients would progress differently and under what conditions is the BEAR group interventions more effective.
Method
Participants
Participants in Study 2 were the same as those in Study 1 (i.e., 307 student clients and 250 teachers).
Measures
Negative Affect. Negative affect was assessed by the Negative Affect (NA) subscale in the international Positive Affect and Negative Affect Short Form (I-PANAS-SF; Thompson, 2007). The NA (5 items) measured the extent to which individuals feel a range of negative affect (e.g., ashamed and upset). Student clients rated items on a 5-point Likert scale (1 = never to 5 = always). Higher scores reflect higher levels of negative affect. We received permission from the original author and used the translation procedure outlined by Æ gisdóttir et al. (2008). The coefficient α was .65 for NA among Singaporean Chinese (Wong et al., 2011). Convergent validity evidence was shown by a negative association between negative affect with both
subjective well-being and happiness (Thompson, 2007). In the present study, coefficient αs for the NA ranged from .65 to .80 (M = .73) across five waves of data.
Emotional Cultivation. The self-report and other report versions of the Emotional Cultivation Scale (ECS; 9-item; Wang, Wei, et al., 2019) were used to measure the knowing (e.g., emotional consequence awareness) and doing (e.g., emotional self-control or flexibly using culturally appropriate emotional regulation strategies) components of the student clients’ emotional regulation process from student
21
clients and teachers’ perspectives. The ECS consisted of two subscales: Understanding Emotion Connotations (UEC; four items) and Cultivating Emotion Strategies (CES; five items). UEC was defined as being aware of current emotions and understanding how one’s thoughts and actions will impact the consequences of emotional responses. CES was defined as developing strategies of regulating emotions through emotional self-control, creating alternative thoughts, and considering the best for self and others.
For better readability, the subscales rated by student clients (i.e., self-report version) were labeled UEC-S and CES-S; the subscales rated by their teachers (i.e., other report version) were labeled UEC-T and CES-T. Sample item for UEC-S was, “I can understand that different ways to express emotions will lead to different
consequences.” Sample item for CES-S was, “When encountering problems, I can change my thoughts to control my temper.” The other report version of the ECS was developed (Wang, Wei, et al., 2019) by replacing the word “I” to “he/she” in the scale items, e.g., “When something happens, I have ways to calm myself down” was changed to “When something happens, he/she has ways to calm herself/himself down.” Student clients and teachers rated each item using a 5-point Likert scale (1 = strongly disagree to 5 = strongly agree), to indicate the extent of what student clients and teachers agreed regarding the student client’s emotional cultivation. Higher scores indicated better emotional cultivation.
Coefficient αs were ranged from .63 to .71 for UEC-S, .74 to .80 for CES-S, .77 for UEC-T, and .86 for CES-T among Taiwanese elementary and middle school
students (Wang, Wei, et al., 2019). Validity was demonstrated by the following
evidence (Wang, Wei, et al., 2019). Convergent validity was supported by the positive associations with cognitive flexibility. Discriminant validity was evidenced by a nonsignificant association with suppression. Concurrent validity was revealed by
22
positive associations with positive affect, basic psychological need satisfaction,
gratitude, responsiveness from teachers and parents. In the present study, coefficient αs were ranged from .68 to .87 (M = .79) for UEC-S, .80 to .87 (M = .83) for CES-S, .80 to .83 (M = .82) for UEC-T, and .86 to .91 (M = .89) for CES-T.
Academic Efficacy. Academic efficacy was assessed by the Academic Efficacy (AE) subscale in the Patterns of Adaptive Learning Scales (PALS; Midgley et al., 2000). The AE (5 items) subscale measured the student clients’ beliefs in their ability to succeed in mastering the classwork. Sample item was, “Even if the work is hard, I can learn it.” Student clients rated each item on a 5-point Likert scale (1 = not at all true to 5 = very true). Higher scores indicated better academic self-efficacy. The coefficient α was .78 among the 5th-grade students (Midgley et al., 2000) and .86 among Taiwanese children and adolescents (Wang, Wei, et al., 2019). Construct validity was demonstrated by positive associations between academic efficacy and task goals in both math and English among the 5th-grade students (Midgley et al., 2000). In the present study, coefficient αs for the AE ranged from .87 to .92 (M = .90) across three waves of data.
Group Intervention
The BEAR group was an 8-session emotional regulation group intervention conducted in the elementary and middle school counseling center. Each session was about 90 minutes. The BEAR group was designed based on the emotional cultivation process (Wang et al., 2012) and dual emotion regulation model (Wang, 2010) in accordance with East Asian cultural characteristics.
Emotional Cultivation Process. Wang and her colleagues (2012) categorized culturally effective and helpful emotion regulation strategies into four BEAR domains (i.e., B: Belief reframing, E: Emotional consequences awareness, A: Action control, R:
Regulatingemotion flexibly through culturally appropriate strategies). This process is
23
called the Emotional Cultivation Process. It was applied to the BEAR group by integrating this theory into the group activities and by converting the four domains of Emotional Cultivation Process into the goals of BEAR group.
To be more specific, student clients were trained to develop awareness of consequences of different emotional responses and ability to cultivate emotions. In situations or stressful events that provoke emotions, student clients learned to first control or hold back the actions that may lead to negative consequences, then identify alternative thoughts, and finally flexibly adopt different appropriate emotion regulation strategies for the particular situation (Wang et al., 2012; Wang, Wei, et al., 2019). These strategies embed the meaning of forbearance in the East Asian cultural context (Wang, Wei, et al., 2019). Meanwhile, they were corresponded with the concept to regulate emotions in socially adaptive ways, which is to adaptively select and vary strategies as needed to fit the context (Gross & Cassidy, 2019).
Dual Emotion Regulation Model. The society in Taiwan retained the
traditional Eastern cultural values, but was also influenced by the Western culture at the same time. Theories and strategies of emotion regulation from the West have long been spread into Taiwan through various means, such as books, lectures, workshops, etc. In results, the development of emotion regulation capabilities of Taiwanese was probably not only affected by the concept of Eastern culture, but also by the concept of Western culture. However, the emotion management framework developed from an
individualistic cultural context might not be completely applicable to the Taiwanese society. To develop an emotion regulation theory for Taiwanese, we need to take cultural considerations (i.e., blend of cultural elements from East and West) into account.
24
Cultural differences exist in the use of emotion regulation strategies (Liddell &
Williams, 2019). Based on the previous studies, expressiveness was most often regarded as adaptive emotion regulation strategy in Western individualism culture (Butler et al., 2009; Kim & Sherman, 2007), while forbearance was taught to be an adaptive emotion regulation strategy in the East Asian culture (Huang et al., 2008; Li & Hsiao, 2008). The maladaptive emotion regulation strategies in Western culture and East Asian culture were suppression (Butler et al., 2007; Soto et al., 2011) and impulsiveness (Eisenberg et al., 2006) respectively.
Wang (2013) categorized these four emotion regulation strategies into two dimensions: (a) contextual sensitivity (appropriate vs. inappropriate in an Asian cultural context), and (b) regulating orientations (regulating outwards vs. regulating inwards).
As seen in Table 4, a two by two framework (i.e., Dual Emotion Regulation Model) with each specific emotion regulation strategy has emerged: (a) Expressiveness (appropriate and regulating outwards), (b) Forbearance (appropriate and regulating inwards), (c) Impulsiveness (inappropriate and regulating outwards), and (d) Suppression (inappropriate and regulating inwards).
Table 4
Dual Emotion Regulation Model
Regulating orientations
Contextual sensitivity Regulating outwards Regulating inwards
Appropriate Expressiveness Forbearance
Inappropriate Impulsiveness Suppression
Note. Adapted from Developing and Validation a Nonwestern Perspective of Emotion Management Model for Children in Taiwan [Paper presentation] by L. Wang, 2013,
25
Society for Psychotherapy Research 44th International Annual Meeting, Brisbane, Australia.
In the BEAR group, the concept of Dual Emotion Regulation Model was introduced to the student clients with four bears: (a) Expressive Bear, (b) Forbearance Bear, (c) Impulsive Bear, and (d) Suppression Bear (Wang, 2013). Expressive Bear is characterized as brave, likes to think, and willing to express. When encountering problems, the Expressive Bear will consider the consequences according to the
situation, then appropriately express the feelings and thoughts. The Expressive Bear is colored in orange, symbolized cheerful. Forbearance Bear is characterized as gentle, optimistic, and does not like conflicts. When encountering problems, the Forbearance Bear can think from different perspectives to regulate emotions. The Forbearance Bear is colored in green, symbolized peaceful. The Impulsive Bear is characterized as brave, straight-forward, and energetic. But sometimes because of carelessness, the Impulsive Bear can get into trouble, and thereafter regret easily. The Impulsive Bear is colored in fiery red. The Suppression Bear is characterized as gentle, shy, and afraid of conflicts.
Therefore, the Suppression Bear will hide emotions or thoughts, feel wronged, and feel not being understood. The Suppression Bear is colored in blue, symbolized suppression.
The difference between the Expressive Bear and the Impulsive Bear lies in whether the consequences after expressing can match according to the expectations of the given social situation, which focused on external consequences. The difference between the Forbearance Bear and the Suppression Bear is the degree of influence on individual mental health after not expressing, which focused on internal consequences.
Group Design Principles. There were several design principles of the BEAR group (Wang et al., 2015). Firstly, the beginning stage of BEAR group focused on
26
building relationships among group members, establishing therapeutic culture in the group, and teaching the student clients about the theories of Emotional Cultivation Process and Dual Emotion Regulation. Secondly, the middle/working and final stage of BEAR group helped the student clients to develop and apply flexible and adaptive emotion regulation skills. Thirdly, to strengthen the effects of situated learning, the opportunities for “here-and-now” experiential learning and practices for actions that can lead to positive consequences were emphasized. These experiences of actual
interpersonal conflicts in the BEAR group, can help student clients to develop the capability to confront and solve problems, and also appropriate ways to control emotions. Fourthly, each session will have “Praise and Appreciation Time”, where student clients complement and thank other group members by pointing out a specific part of what they have done or said throughout the session.
Finally, the Counseling Communication Log was used as the homework. The implementation of Counseling Communication Log involved a weekly routine where student clients wrote down what they have learned at the end of each session, and counselors (i.e., the group leaders) provided feedback to student clients. Then, student clients will take back the Counseling Communication Log and complete a task that is designed according to the goal of each session. Student clients returned the Counseling Communication Log to counselors before the next group session.
Procedure
The data of 307 student clients were retrieved from the database of Dual
Emotion Regulation Research Project. All of the student clients voluntarily participated in 53 BEAR groups across three semesters (i.e., Fall 2015, Spring 2017, and Fall 2017).
Because of the different strategies of data collection in each semester, the number of waves of data collection for each variable (i.e., NA, UEC-S, CES-S, UEC-T, CES-T,
27
and AE) in each semester were different (see Table 5). For the analysis in Study 2, NA, UEC-S, and CES-S have five waves of data; UEC-T, CES-T, and AE have three waves of data. For the five waves data, the five time points of data collections were: Week 0 (a week before student clients started to join the first session of BEAR group), Week 3 (after the third session), Week 6 (after the sixth session), Week 9 (a week after the eighth session [i.e., the last session of BEAR group]), and Week 13 (five weeks after the last session). For the three waves data, the three time points of data collections were:
Week 0 (equivalent to pretest), Week 9 (equivalent to posttest), and Week 13
(equivalent to follow-up). All the scales used in this study were presented in Traditional Chinese.
Table 5
Numbers of Waves of Data Collection for Three Semesters Semester n No. of
groups
No. of waves of data collection
NA UEC-S CES-S UEC-T CES-T AE
Fall 2015 159 28 5 3 3 3 3 3
Spring 2017 84 14 5 5 5 3 3 3
Fall 2017 64 11 0 5 5 3 3 3
Note. NA = Negative Affect; UEC-S = Understanding Emotion Connotations from Student Clients’ Perspectives; CES-S = Cultivating Emotion Strategies from Student Clients’ Perspectives; UEC-T = Understanding Emotion Connotations from Teachers’
Perspectives; CES-T = Cultivating Emotion Strategies from Teachers’ Perspectives; AE
= Academic Efficacy.
28 Data Analysis
In Study 2, multilevel modeling (MLM; via hierarchical linear modeling [HLM 8.00 Student version]) was used to analyze the data. A 3-level HLM analysis was performed in this study. Level-1 analysis corresponded to between-sessions, Level-2 analysis corresponded to between-student client, and Level-3 analysis corresponded to between-group. As student clients were nested within individual and within groups, using a traditional regression method to analyze data at the individual level would omit the dependence in data within groups (Cheng et al., 2008). Thus, choosing MLM as the analytic strategy was more appropriate for assessing and handling nested data.
To make comparisons, dummy variables were created to examine the effects of student client types and group composition. The student clients with internalizing problems was used as the reference group for Level-2. At The percentage of student clients with internalizing problems was used as the reference group for Level-3.
Initially, a completely unconditional 3-level HLM analysis was run to partition the variance in NA scores into group, student client, and between-session components. For NA, 53% of the variance was between between-sessions, 47% of the variance was between student clients (intraclass correlation coefficient [ICC] = .47, χ2(df = 201) = 1080.99, p < .001), and 0% of the variance was between groups (ICC
= .00, χ2(df = 41) = 43.45, p = .367). Therefore, there was sufficient variance in NA scores at session, student client, but not group levels.
A completely unconditional 3-level HLM analysis was run to partition the variance in UEC-S scores into group, student client, and between-session components. For UEC-S, 54% of the variance was between between-sessions, 32% of the variance was between student clients (ICC = .32, χ2(df = 254) = 834.96, p < .001), and 14% of the variance was between groups (ICC = .14, χ2(df = 52) = 141.03, p < .001).
29
Therefore, there was sufficient variance in UEC-S scores at session, student client, and group levels.
A completely unconditional 3-level HLM analysis was run to partition the variance in CES-S scores into group, student client, and between-session components. For CES-S, 62% of the variance was between between-sessions, 33% of the variance was between student clients (ICC = .33, χ2(df = 254) = 786.16, p < .001), and 5% of the variance was between groups (ICC = .05, χ2(df = 52) = 86.56, p = .002).
Therefore, there was sufficient variance in ECS-CES scores at session, student client, and group levels.
A completely unconditional 3-level HLM analysis was run to partition the variance in UEC-T scores into group, student client, and between-session components. For UEC-T, 62% of the variance was between between-sessions, 33% of the variance was between student clients (ICC = .33, χ2(df = 253) = 643.73, p < .001), and 6% of the variance was between groups (ICC = .06, χ2(df = 52) = 86.28, p = .002).
Therefore, there was sufficient variance in UEC-T scores at session, student client, and group levels.
A completely unconditional 3-level HLM analysis was run to partition the variance in CES-T scores into group, student client, and between-session components. For CES-T, 49% of the variance was between between-sessions, 43% of the variance was between student clients (ICC = .43, χ2(df = 253) = 906.79, p < .001), and 9% of the variance was between groups (ICC = .09, χ2(df = 52) = 97.98, p < .001).
Therefore, there was sufficient variance in CES-T scores at session, student client, and group levels.
A completely unconditional 3-level HLM analysis was run to partition the variance in AE scores into between-group, between-student client, and between-session
30
components. For AE, 44% of the variance was between sessions, 50% of the variance was between student clients (ICC = .50, χ2(df = 254) = 1094.46, p < .001), and 6% of the variance was between groups (ICC = .06, χ2(df = 52) = 82.50, p = .005). Therefore, there was sufficient variance in AE scores at session, student client, and group levels.
In Model 1, we examined how NA, UEC-S, CES-S, UEC-T, CES-T, and AE changed across weeks. Specifically, we conducted an unconditional growth model with week predicting NA, UEC-S, CES-S, UEC-T, CES-T, and AE, respectively. Below is an example of one of these unconditional growth models.
Level-1 Model.
NA = π0 + π1*(WEEK) + e Level-2 Model.
π0 = β00 + r0
π1 = β10 + r1
At level 2 and Level-3 Model.
β00 = γ000 + u00
β10 = γ100 + u10
In Model 2, conditional growth models, we examined how types of student clients at level-2 and compositions of types of student clients within groups at level-3 related to growth in NA, UEC-S, CES-S, UEC-T, CES-T, and AE, respectively. Below is an example of one of these conditional growth models.
Level-1 Model.
NA = π0 + π1*(WEEK) + e Level-2 Model.
π0 = β00 + β01*(HIDDEN Type) + β02*(IMPULSIVE Type) + r0
π1 = β10 + β11*(HIDDEN Type) + β12*(IMPULSIVE Type) + r1
31 Level-3 Model.
β00 = γ000 + γ001(Percentage of HIDDEN Type) + γ002(Percentage of IMPULSIVE Type) + u00
β01 = γ010 + u01
β02 = γ020 + u02
β10 = γ100 + γ101(Percentage of HIDDEN Type) + γ102(Percentage of IMPULSIVE Type) + u10
β11 = γ110 + u11
β12 = γ120 + u12
Results
Descriptive Analyses
Table 6 displays the descriptive statistics (mean and standard deviation) for each variable at each time point of data collections. The missing data in Study 2 were
handled with pairwise deletion, all available data were used by eliminating any cases with missing data on an analysis-by-analysis basis (Peugh & Enders, 2004). Multilevel data are often incomplete (Grund et al., 2019), the pairwise deletion was implemented to maximize sample size by not requiring complete data on all variables, under the
assumption that missing data are missing due to a missing completely at random mechanism in this study.
32 Table 6
Descriptive Statistics for Study Variables
Variable Week 0 Week 3 Week 6 Week 9 Week 13
n M (SD) n M (SD) n M (SD) n M (SD) n M (SD)
NA 242 2.02 (0.76) 242 2.03 (0.79) 239 2.01 (0.82) 236 1.91 (0.84) 237 1.97 (0.86) UEC-S 306 3.71 (0.77) 146 4.16 (0.77) 144 4.18 (0.78) 301 4.00 (0.81) 299 3.95 (0.88) CES-S 307 3.24 (0.86) 146 3.73 (0.81) 144 3.71 (0.89) 301 3.66 (0.82) 299 3.67 (0.85)
UEC-T 305 3.26 (0.68) — — — — 300 3.63 (0.61) 300 3.74 (0.59)
CES-T 305 2.77 (0.76) — — — — 300 3.26 (0.79) 300 3.28 (0.78)
AE 307 3.41 (0.98) — — — — 301 3.60 (1.02) 300 3.63 (1.00)
Note. NA = Negative Affect; UEC-S = Understanding Emotion Connotations from Student Clients’ Perspectives; CES-S = Cultivating Emotion Strategies from Student Clients’ Perspectives; UEC-T = Understanding Emotion Connotations from Teachers’ Perspectives; CES-T = Cultivating Emotion Strategies from CES-Teachers’ Perspectives; AE = Academic Efficacy.
33 Multilevel Analyses
Negative Affect. In the unconditional growth model for NA, the fixed effect for weeks was not significant (γ100 = −0.01, p = .115). Therefore, student clients’ negative affect did not change, significantly over time. The random effects for the unconditional growth model for NA showed that the growth coefficient differed significantly between student clients (r1 = .002, p < .001) but not between groups (u10 < .001, p = .395).
In the three-level conditional growth model for the final estimation of the fixed effects (see Table 7), the significant intercept for NA is trivial because the scoring of NA does not contain “0”. For the intercept, there is a significant effect on the percentage of impulsive student clients in a group (γ002 = −1.04, p = .003). This shows that student clients, in groups with a larger proportion of student clients with impulsive problems, had a lower initial level of negative affect. For the intercept, there is also a significant effect for student clients with hidden problems (γ010 = 0.34, p = .001). This shows that student clients with hidden problems had a higher, initial level of negative affect. For the intercept, there is also a significant effect for student clients with impulsive problems (γ020 = 0.32, p = .006). This shows that student clients with impulsive problems had a higher, initial level of negative affect.
The effect for weeks was significant (γ100 = −0.03, p = .027). This shows that the student clients in the BEAR groups decreased their negative affect by .03 points per week. The between-level interaction for the percentage of student clients with hidden problems and change in NA was significant (γ101 = 0.05, p = .021). This significant interaction is displayed in Figure 2. As seen in Figure 2, when there are more student clients with hidden problems the negative affect of the student clients does not change over time (simple slope = .007, p = .536); however, when there are fewer student clients with hidden problems in a group, negative affect decreased marginally significant over
34
time (simple slope = −.016, p = .087). This shows that the BEAR groups are more effective in decreasing the negative affect when the group has fewer student clients with hidden problems.
Figure 2
Between-Level Interaction for the Percentage of Student Clients with Hidden Problems and Change in Negative Affect
The between-level interaction for the percentage of student clients with impulsive problems and change in NA was significant (γ102 = 0.09, p = .017). This significant interaction is displayed in Figure 3. As seen in Figure 3, when there are more student clients with impulsive problems the negative affect of the student clients does not change over time (simple slope = −.001, p = .937); however, when there are fewer student clients with impulsive problems in a group, negative affect decreased
1.0
35
significantly over time (simple slope = −.023, p = .032). This shows that the BEAR groups are more effective in decreasing negative affect when the group has fewer student clients with impulsive problems.
Figure 3
Between-Level Interaction for the Percentage of Student Clients with Impulsive
Between-Level Interaction for the Percentage of Student Clients with Impulsive