Researchers of positive psychology have highlighted the importance of individuals’ well-being and engagement at the same time (Seligman, 2011). Learning engagement is a key factor in students’ academic learning outcomes (Moreira et al., 2015); however, students with special educational needs (SENs) usually experience a lower degree of engagement than do students without SENs (Rangvid, 2018). Students with SENs may encounter negative developmental experiences of numerous types (Moreira et al., 2015), such as lower learning performance, perceived disadvantages, and greater negative feelings toward learning (Cheng & Chang, 2014). Yu (2015) determined that individuals who did not exhibit high performance and frequently felt helpless when learning may have had decreased well-being. Numerous studies have reported that regular-class students’ learning engagement affects their well-being and learning engagement can also be predicted by individuals’ wellbeing. Moreover, learning engagement and well-being may interact with each other. In Taiwan, approximately 80% of middle school students with SENs receive special education in resource rooms. However, relatively few studies explore the learning engagement and well-being of students with SENs. This study examined the relationship between the learning engagement and well-being of junior high school students learning in resource rooms.

This study used a questionnaire to collect information on 143 junior high school students with SENs in Hsinchu, Taiwan, for a pilot study. In the formal survey study, 324 (32% girls and 68% boys) junior high school (24% seventh grade, 35% eighth grade, and 41% ninth grade) students with SENs (64% learning disabilities, 9% emotional or behavior disorders, 9% intellectual

disabilities, 14% autism, 1% hearing impairments, and 3% other disabilities) were recruited to examine their learning engagement, well-being, and the relationship between the two variables.

The participants rated each item on the self-developed questionnaire on a 5-point Likert scale (5 = completely agree, 4 = strongly agree, 3 = agree, 2 = disagree, 1 = strongly disagree); a higher total score indicated a higher level of learning engagement and well-being. Exploratory factor analysis (EFA) was conducted with 143 pilot participants and confirmatory factor analysis (CFA) was conducted with 324 participants. Information regarding each scale used is presented as follows:

The learning engagement scale was developed on the basis of the four aspects of learning engagement reported by Reeve and Tseng (2011). In addition, the Engagement and Disengagement Scale (Jang et al., 2016) was adapted to evaluate the learning engagement of junior high school students with SENs who were learned in resource rooms. A Kaiser–Meyer–Olkin (KMO) and Bartlett’s tests exhibited a high KMO score (.921) and all items exhibited significant differences (χ2 = 1909.453, p < .000). The EFA demonstrated that the final version of the scale had four factors, which together accounted for 74.12% of the variance. The first factor agentic engagement included five items and exhibited an internal consistency of .89. The second factor behavioral engagement included four items and exhibited an internal consistency of .89. The third factor cognitive engagement included five items and exhibited an internal consistency of .90. The fourth factor emotional engagement included four items and exhibited an internal consistency of .86. The total scale’s Cronbach’s α value was .94. CFA was used to evaluate the fit of the four-factor structure with data from 324 Taiwan junior high school students with SENs. The learning engagement model was a good fit for the data (χ2/df = 1.939, goodness of fit index [GFI] = .922, adjusted GFI [AGFI] = .898, comparative fit index [CFI] = .955, root mean square error of approximation [RMSEA] = .054, and standardized root mean squared residual [SRMR] = .043). The composite reliability (CR) values of agentic, behavioral, cognitive, and emotional engagement were .816, .806, .857, and .848, respectively. The average variance extracted (AVE) values of these variables were .473, .513, .546, and .584, respectively. This scale had high convergent validity.

The Well-Being Scale was developed on the basis of the three aspects of subjective well-being (emotional, psychological, and social well-being) reported by Keyes and Magyar-Moe (2003) and adapted from the Well-Being Scale (Jhang, 2019) and the Children՚s Well-Being Scale (Chiou, 2009) to evaluate the well-being of junior high school students with SENs. The CFA yielded a high KMO score (.933), and all items exhibited significant differences (χ2 = 1163.561, p < .000). The final scale had three factors, which together accounted for 68.12% of the variance. The first factor social well-being included six items and exhibited an internal consistency of .89. The second factor emotional well-being included three items and exhibited an internal consistency of .79. The third factor psychological well-being included five items and exhibited an internal consistency of .84. The total scale’s Cronbach’s α value was .93. We used CFA to evaluate the fit of the three-factor structure with data from the 324 Taiwan junior high school students with SENs. The well-being measurement model exhibited an acceptable fit for the data (χ2/df = 2.222, GFI = .935, AGFI = .908, CFI = .955, RMSEA = .062, SRMR = .041). The CR values of social, emotional, and psychological well-being, were .852, .765, and .820, respectively. The AVE of these variables were .490, .521, and .479, respectively. This scale exhibited high convergent validity.

To examine discriminant validity, we used the square root of the latent variables’ AVE value. If the square root of each AVE value was considerably greater than any correlation among any pair of latent constructs, the discriminant validity could be assured. The latent variables of this study exhibited acceptable discriminant validity; thus, the dimensions of the learning engagement or well-being scales developed could be differentiated.

Multivariate analysis of variance revealed no difference in students’ learning engagement on the basis of sex or grades in junior high school students with SENs in resource rooms, but individual students’ learning engagement was affected by their type of disability. Students with intellectual disabilities exhibited higher emotional engagement in resource rooms than did students with other disabilities, and students with intellectual disabilities exhibited higher agentic engagement in resource rooms than did students with learning disabilities. By contrast, students’ well-being did not differ by grade or disability type. However, boys perceived greater emotional well-being than did girls.

Correlation and path analysis demonstrated that learning engagement and well-being were highly positively correlated and students’ learning engagement in resource rooms and well-being were mutual predictors. Emotional and agentic engagement affected individuals’ well-being, and psychological well-being affected learning engagement.

On the basis of the results, the following suggestions are provided. First, for resource room teachers and school administrators, creating a positive learning environment, guiding students to express their learning needs, and integrating a learning strategy into academic subjects can promote students’ well-being. Second, highlighting or exploring individual students’ skills to enhance their well-being may increase the learning engagement of students with SENs. Future research can investigate the effects of students’ cognitive ability and the frequency of resource room service received. Mediator variables should be examined to better demonstrate the relationship between learning engagement and well-being in students with SENs in Taiwan.