Research in Developmental Disabilities 36 (2015) 45–54
Contents lists available at ScienceDirect
Research in Developmental Disabilities
The relationship between self-determination and academic achievement for adolescents with intellectual disabilities Amy S. Gaumer Erickson *, Patricia M. Noonan 1, Chunmei Zheng, Jennifer A. Brussow 2 University of Kansas, Center for Research on Learning, 521 JR Pearson Hall, 1122 West Campus Road, Lawrence, KS 66045, United States
A R T I C L E I N F O
A B S T R A C T
Article history: Received 2 May 2013 Received in revised form 8 September 2014 Accepted 9 September 2014 Available online 11 October 2014
Previous research has demonstrated that for students with intellectual disabilities, improved self-determination skills are positively correlated with productivity and organization during school and quality of life outcomes in adulthood. Despite extensive investigation in these areas, the predictive relationship between self-determination and academic achievement for students with intellectual disabilities has not been fully established. This study utilized the sample from the National Longitudinal Transition Study-2 of 480 adolescents with intellectual disabilities in the United States in an attempt to provide a possible empirical explanation of the relationship between academic achievement and self-determination, taking into account the covariates of gender, family income and urbanicity. The structural equation model was found to closely fit the data: all path coefficients were statistically significant. The results of this study identify a strong correlation between self-determination and academic achievement for adolescents with intellectual disabilities, indicating a linear relationship of these skills and supporting an increased focus on the teaching of self-determination skills. ß 2014 Elsevier Ltd. All rights reserved.
Keywords: Mental retardation Developmental disabilities Intellectual disabilities Self-determination Academic achievement
1. Introduction Academic achievement, typically defined as proficiency in reading and mathematics, has consistently been identified as a predictor of post-school success, including social inclusion, economic self-sufficiency, and overall quality of life (Day & Newburger, 2002; Greene, 2000; Kutner et al., 2007; O’Neill, 2001; Swanson, 2004). Moreover, research has explicitly related higher levels of achievement in reading, writing, math, and problem-solving skills with improved post-school outcomes for students both with and without disabilities (Benz, Yovanoff, & Doren, 1997; Schneider, Kirst, & Hess, 2003). In order to secure these positive post-school outcomes for students, educational policy and research continues to focus on promoting academic achievement. Scholarship in post-school outcomes for students with intellectual disabilities has identified self-determination as an important attribute. Specifically, researchers have documented (a) the efficacy of self-determination interventions for
* Corresponding author. Tel.: +1 785 864 0517. E-mail addresses: [email protected] (A.S. Gaumer Erickson), [email protected] (P.M. Noonan), [email protected] (C. Zheng), [email protected] (J.A. Brussow). 1 Tel.: +1 785 864 0593. 2 Tel.: +1 785 864 0517. http://dx.doi.org/10.1016/j.ridd.2014.09.008 0891-4222/ß 2014 Elsevier Ltd. All rights reserved.
46
A.S. Gaumer Erickson et al. / Research in Developmental Disabilities 36 (2015) 45–54
students with intellectual disabilities, (b) a positive relationship between self-determination and post-school outcomes for students with intellectual disabilities, and (c) a positive relationship between self-determination and access to the general education curriculum for students with intellectual disabilities (Agran, Blanchard, & Wehmeyer, 2000; Agran, Blanchard, Wehmeyer, & Hughes, 2001; Brooks, Todd, Tofflemoyer, & Horner, 2003; Craft, Alber, & Heward, 1998; Hughes et al., 2002; McCarl, Svobodny, & Beare, 1991; Moes, 1998; O’Reilly, Lancioni, Gardiner, Tiernan, & Lacy, 2002; Rock, 2005; Warner & de Jung, 1971). Furthermore, self-determination has been shown to be positively correlated with school engagement and adult outcomes for students with intellectual disabilities (Agran et al., 2005; Arndt, Konrad, & Test, 2006; Sa´nchez & Roda, 2003; Wehmeyer & Palmer, 2003; Wehmeyer & Schalock, 2001; Wehmeyer & Schwartz, 1997, 1998). Despite this body of research, studies overall exhibited a ‘‘lack of focus on academic skills in the self-determination literature for students with mental retardation/developmental delays’’ (Fowler, Konrad, Walker, Test, & Wood, 2007, p. 281). The current study incorporated structural equation modeling to investigate the direct relationship between self-determination and academic achievement for students with intellectual disabilities. 2. Theory Working within theories of human agentic behavior, Wehmeyer and Little (2009) describe a self-determined person as the ‘‘origin of his or her actions, [who] has high aspirations, perseveres in the face of obstacles, sees more and varied options for action, learns from failures, and overall, has a greater sense of well-being’’ (p. 868). Wehmeyer (2005) has also defined self-determined behavior as ‘‘volitional, intentional, and self-caused, or self-initiated action’’ (p. 115). Self-determined behavior differs from ‘‘other-determined behavior’’ by specifying self as the primary causal agent as opposed to behavior caused by external forces (Wehmeyer, Abery, et al., 2011). Self-determined behavior is affected by environmental factors as well as by individual knowledge, values, and skills (Field & Hoffman, 2001, 2007; Hoffman & Field, 2005). The influence of self-determination interventions on students with disabilities has been the focus of substantial research. Researchers have explored various methods to increase self-determination, including evidence-based practices that use choice-making, goal-setting, and self-advocacy to increase students’ self-determined behavior (Ezell, Klein, & Ezell-Powell, 1999; Fullerton & Coyne, 1999; Martin et al., 2003; Wehmeyer, Palmer, Lee, Williams-Diehm, & Shogren, 2011). Findings indicate that interventions are effective for improving self-determination (Chambers et al., 2007). Research has also examined the impact of self-determination skills on class participation and direction-following skills (Agran et al., 2005; Fowler et al., 2007; Konrad, Fowler, Walker, Test, & Wood, 2007; Lee, Wehmeyer, Soukup, & Palmer, 2010). Studies have used a variety of methods to assess the effect of self-determination interventions on quality and productivity in language arts and math; however, the existing research has focused primarily on skills such as organization and productivity rather than on academic achievement outcomes (Fowler et al., 2007). It is documented in the academic literature that individuals with developmental or intellectual disabilities are less likely to exhibit self-determined behavior due to their limited choice-making opportunities (Wehmeyer, Kelchner, & Richards, 1995, 1996; Wehmeyer & Metzler, 1995). As reported by Wehmeyer and Palmer (2003) and Wehmeyer and Schwartz (1997), higher self-determination scores were correlated with more positive post-school outcomes for students with intellectual disabilities. Additionally, Wehmeyer and Schwartz (1998) showed that self-determination skills corresponded with a positive quality of life for people with intellectual disabilities; this finding was replicated in an international survey of adults with mild intellectual disabilities (Lachapelle et al., 2005). Moreover, though it is a common assumption that cognitive limitations result in limitations in self-determination skills (Wehmeyer, Abery, et al., 2011), research has demonstrated that the opportunity to make choices was the most significant variable predicting levels of self-determination (Wehmeyer & Garner, 2003; Wehmeyer, Palmer, et al., 2011). Many moderating variables such as culture, gender, age, cognitive ability, religious beliefs, environment, and experiences of oppression have all been shown to impact self-determination (Duvdevany, Ben-Zur, & Ambar, 2002; Kurtz-Costes, Rowley, Harris-Britt, & Woods, 2008; Nota, Ferrari, Soresi, & Wehmeyer, 2007; Shogren et al., 2007; Wehmeyer, Abery, et al., 2011; Wehmeyer & Bolding, 1999, 2001; Wehmeyer, Palmer, et al., 2011). Student achievement is also influenced by moderating factors including socioeconomic status, gender, and urbanicity. For example, research has found that family and community poverty are frequently associated with negative effects on academic achievement (Blackorby, Chorost, Garza, & Guzman, 2004; Hattie, 2009; Sharkey, 2009; Wagner, Newman, Cameto, & Levine, 2006). The National Longitudinal Transition Study-2 (NLTS2) classifies additional covariates as demographic: age, ethnicity, family (e.g., support, expectations), and school characteristics and experiences (e.g., grade retention, absenteeism, behavior at school) (Wagner et al., 2006). The complex nature of the interrelationships between factors complicates research on the relationship between self-determination and academic achievement for students with intellectual disabilities. To advance scholarship on the relationship between self-determination and academic achievement, this study applied a structural equation model developed by the authors (Zheng, Gaumer Erickson, Kingston, & Noonan, 2014) that found a positive linear relationship between self-determination and academic achievement for adolescents with learning disabilities. To test the hypothesized linearity of the variables for adolescents with intellectual disabilities, two research questions were investigated: A. Is there a direct relationship between self-determination and academic achievement for adolescents with intellectual disabilities?
A.S. Gaumer Erickson et al. / Research in Developmental Disabilities 36 (2015) 45–54
47
B. Is self-determination a good predictor of academic achievement for adolescents with intellectual disabilities?
3. Methods 3.1. Participants This study is based on the data from Waves 1 and 2 of the National Longitudinal Transition Study-2 (NLTS2). NLTS2 was commissioned by the U.S. Department of Education, Office of Special Education Program (OSEP) in order to catalog the home and school experiences, academic performance, and post-school outcomes of a national sample of students with disabilities over a period of 10 years. This longitudinal data collection began in 2002 with Wave 1, which collected data on 11,000 youth across the United States aged 13–16, and concluded with Wave 5, which followed up with these same individuals at ages 23–26. Wave 1 and Wave 2 included surveys of parents, youth, and school staff as well as assessments of reading and mathematics achievement, self-determination, self-concept, and friendship. The direct assessments within these waves, which constitute the dataset for the current study, were conducted in 2002 and 2004 with youth 16–18 years of age. NLTS2 data mirrored the 12 disability categories (i.e., learning disabilities, speech impairment, mental retardation, emotional disturbance, hearing impairment, visual impairment, orthopedic impairment, other health impairment, autism, traumatic brain injury, multiple disabilities, and deaf/blindness) outlined in the Individuals With Disabilities Education Improvement Act (IDEA, 2004). The current study focused exclusively on participants identified with a primary disability of mental retardation reported on the students’ individualized education programs (IEPs), yielding a total sample size of 480 students. This sample is referred to as students with intellectual disabilities within the current study. The focus on a single disability category was purposeful, making these results more specific and descriptive than a combination of disability categories would allow. In line with national population trends, the dataset of youth with intellectual disabilities included 52% males, 54.8% white, 33.3% African American and 9.6% Hispanic students. The youth with intellectual disabilities represented individuals in low-income families (54.9%), single-parent households (34.5%), households that did not use primarily English at home (6.3%), and households with a head of household who did not complete high school (32.3%) (Wagner et al., 2003). 3.2. Measures & administration procedures The NLTS2 direct assessment data collection spanned six direct assessment domains: reading, math, content (i.e., science and social studies), self-concept, self-determination, and friendship interaction. The analyses within the current study focused on three of the six domains (reading, math and self-determination). These assessments were administered by trained personnel, typically school psychologists or teachers, who were hired and supervised by NLTS2 staff. Assessments were generally conducted at the adolescent’s school during non-instructional time. All instrument protocols were ‘‘fully scripted to maintain consistency of administration across assessors’’ (Wagner et al., 2006, p. 11). NLTS2 booklets were used to record the responses of each student and then sent to NLTS2 program staff for scoring (Wagner et al., 2006). 3.2.1. Academic achievement measure The NLTS2 assessors administered the Woodcock-Johnson Research Edition (Woodcock, McGrew, & Mather, 2007) to measure students’ academic achievement in math and reading. This research edition is a shortened form of the published Woodcock-Johnson III; its average reliability is .65 (Wagner et al., 2006). Though the published version contains additional items, both versions measure the same constructs and report on the same score scales. Two indicators for math (i.e., applied problems and calculation) and reading (i.e., synonyms/antonyms and passage comprehension), were calculated by NLTS2 staff (Wagner et al., 2006). Reported scores for math and reading were converted to z-scores within the current study. 3.2.2. Self-determination measure The Arc Self-Determination Scale measures individual differences in students’ tendencies to function in a self-determined way (Wehmeyer, 1995). The 72-item measure, with items categorized into one of four subdomains (i.e., autonomy,self-regulation, self-realization, and psychological empowerment) was normed on a sample of 500 youth, primarily identified as having intellectual and learning disabilities, reporting a coefficient alpha of .90. The NLTS2 researchers modified the Arc Self-Determination Scale by reducing the number of items in each domain and deleting the subdomain of self-regulation (Facts from OSEP’s National Longitudinal Studies, 2005). The modified scale contained 26 items across three subdomains: autonomy (15 items), self-realization (five items), and empowerment (six items). Autonomy was measured by a four-point scale: (1) not when I have the chance, (2) sometimes, (3) most of the time, and (4) every time I have the chance. A four-point scale was also used for self-realization questions: (1) never agree; (2) sometimes agree; (3) usually agree; and (4) always agree. Empowerment items utilized a two-point scale. Of these items, one represented passive and pessimistic actions (e.g., ‘‘It is no use to keep trying because it will not change things’’) and the second represented active and optimistic actions (e.g., ‘‘I keep trying even after I get something wrong’’).
48
A.S. Gaumer Erickson et al. / Research in Developmental Disabilities 36 (2015) 45–54
3.3. Analysis The structural equation model incorporated into the data analysis was replicated from a model that utilized a NLTS2 sample of students with learning disabilities (Zheng et al., 2014). The results of the previous study showed a significant positive relationship between self-determination and academic achievement for students with learning disabilities. To test the validity of the model, the same structural equation methodology was applied to the NLTS2 sample of students with mental retardation as their primary IDEA disability classification (referred to as students with intellectual disabilities within the current study). Because the scales varied within the assessments, all scores were converted to z-scores for analysis. Research questions were tested via structural equation modeling following four steps: (1) first-order confirmatory factor analysis (CFA) to specify the relationship between observed and latent variables and the relationships among the latent variables; (2) second-order CFA with beta coefficients to test the significance of these relationships; (3) latent regression between self-determination and academic achievement to test the linearity of the relationship; and (4) inclusion of the covariates gender, family income, and urbanicity in the beta coefficients to test the significance of the overall model. The first-order CFA included five variables: math (M), reading (R), autonomy (AU), self-realization (SR), and empowerment (EM). In the second-order CFA, academic achievement (AA) included the math and reading as subdomains, and self-determination (SD) included autonomy, self-realization, and empowerment as subdomains. The covariates were coded into categories in this data set. ‘‘Male’’ served as the baseline for the gender comparison (Gen); thus, Gender 1 (Gen1) was the female group. Family incomes below $25,000 constituted the baseline group for income (Inc); Income 1 (Inc1) referred to family incomes between $25,000 and $50,000, and Income 2 (Inc2) referred to family incomes over $50,000. Youth from rural areas were the baseline group for urbanicity (Urb), which resulted in Urbanicity 1 (Urb1) referring to suburban areas and Urbanicity 2 (Urb2) referring to urban areas. The three variables of autonomy, self-realization, and empowerment were parceled to form manifest indicators for each of the latent variables. A parcel is defined as ‘‘an aggregate-level indicator comprised of the sum or average of two or more items, responses, or behaviors’’ (Little, Cunningham, Shahar, & Widaman, 2002, p. 152). The advantages of parceling over item-level modeling include higher reliability, small and equal intervals, and a decreased chance of distributional violations. An item-to-variable balancing technique was used to create the parcels for the three variables. Item-total correlations were computed using SPSS, and items of higher correlations were then combined with items of lower correlations. For the five items in the self-realization domain, items 1 and 5 were parceled as the first indicator, items 2 and 4 as the second indicator, and item 3 was the third indicator. Since the scales of measurement varied widely across variables, each indicator was standardized across all variables to have a mean of 0 and a standard deviation of 1.0. The variance was specified as 1.0 for both the first- and second-order variables in order to establish common scales for all variables. This process allowed the variance/covariance matrix of all variables to produce a correlation matrix. Factor loading equations were used to improve the model identification because there were only two items for both the math and reading indicators. 3.4. Missing data The data set had a very small proportion of missing data, on average 3.5% for each variable. In an effort to improve the accuracy of statistical analyses, the multiple-imputation method was performed to impute the missing data and obtain a complete data set. The proportion of missing data was small enough that only one repetition of the multiple imputation method was necessary. All variables (e.g., math score, reading score, gender, and family income) were considered for the imputation in order to minimize bias and increase the accuracy of parameter estimates. 3.5. Weights The stratified and clustered sampling design of the NLTS2 provided a robust dataset; however, this design incorporated proportional oversampling and undersampling of some population subgroups, thus both sampling and replicative weights were provided in the original dataset. Because clustered sampling can result in case similarities, replicative weights were included in data analyses. These replicative weights were intended to reduce the estimate bias and account for the independence of observed cases. Additionally, sampling weights ensured that analyses accurately represented the population of students with intellectual disabilities in the United States. 4. Results Sample sizes have been rounded to the nearest 10 for all results as per the IES restricted-use data requirements of disclosure protection (Institute for Education Sciences, 2012). For the 480 youth with intellectual disabilities, the average math and reading z-scores were 0.74 (SD = 0.81) and 0.86 (SD = 0.78), respectively. The 420 respondents’ average ratings on each of the self-determination subscales were skewed toward positive options. For example, the two-point empowerment scale had a mean of 1.80 (SD = 0.19) indicating that most students viewed themselves as more empowered than not. On the four-point self-realization scale, the mean of 3.08 (SD = 0.55), indicates that most students considered themselves to be highly self-regulated. The four-point scale for the autonomy domain yielded a mean of 2.93 (SD = 0.47), suggesting that most students believed they typically acted autonomously.
A.S. Gaumer Erickson et al. / Research in Developmental Disabilities 36 (2015) 45–54
49
Table 1 Correlations among the academic achievement and self-determination variables for adolescents with intellectual disabilities. Construct
Academic achievement
Autonomy
Self-realization
Empowerment
Academic achievement Autonomy Self-realization Empowerment
1.000 0.208 0.286 0.553
1.000 0.738 0.505
1.000 0.627
1.000
4.1. Factor analysis The factor analysis identified positive correlations between academic achievement and each of the self-determination subscales (see Table 1). The three self-determination subscales demonstrated strong correlations, indicating a clear pattern among the first-order variables. Furthermore, each self-determination subdomain was found to correlate positively with academic achievement (i.e., combined reading and mathematics proficiency). These correlations imply that students with higher levels of autonomy, self-realization, and/or empowerment exhibit higher levels of academic achievement. All factor loadings were significant. The reading, mathematics, and each of the three self-determination subscales all had factor loadings significant at the .01 level, and each variable’s z-score was above the critical value of 1.96. The standardized Beta results represent each variable’s standardized factor loadings. After controlling for gender, family income, and urbanicity, all standardized Beta statistics except for self-realization resulted in significant p values at the .01 confidence level (see Table 2). The self-realization subscale seemed to have been substantially influenced by the covariate variables, which may explain its exceptionality. Overall, factor loadings indicate that the variables of interest were strongly related. 4.2. Structural model To depict the relationship between academic achievement and self-determination after controlling for gender, family income, and urbanicity, a model with these variables as covariates was generated (see Fig. 1). The analysis found self-determination to be a significant predictor for academic achievement even after controlling for the covariant variables (b = 0.380, p = .001), as shown in Table 3. Urbanicity was also a significant predictor for academic achievement, with students attending urban schools having significantly lower levels of academic achievement. The results did not identify family income or gender as significant predictors of academic achievement for adolescents with intellectual disabilities. Although one of the three covariates was found to be a significant predictor of academic achievement, self-determination remained the strongest predictor of academic achievement even when controlling for gender, family income and urbanicity (beta = 0.350, p = .001). Interpretation of the standardized regression coefficient (Beta ST) reveals that an increase of one standard deviation in the level of self-determination results in an increase of .35 standard deviation for achievement in reading and mathematics for students with intellectual disabilities. This finding supports a direct relationship between self-determination and academic achievement for adolescents with intellectual disabilities. The structural equation model showing the second-order factor loadings and the overall standardized Beta statistics after controlling for gender, family income, and urbanicity can be found in Fig. 1. 4.3. Limitations This study’s primary limitation can be found in the measures themselves. NLTS2 only incorporated selected items from the self-determination scale (i.e., 26 of the 72 items in the Arc Self-Determination Scale). Additionally, an entire subscale, self-regulation, was not included in the NLTS2 data collection. Given this limitation, we cannot predict whether the linear relationship identified by this analysis would have been consistent if the full version of the Arc Self-Determination Scale had been administered. Table 2 Structural equation model: factor loadings after adding covariate variables. Construct
Factor loadings (SE)
Self-determination Autonomy Self-realization Empowerment
0.956 (0.152) 5.205 (2.115) 1.218 (0.322)
Academic achievement Math: Applied problems Math: Calculations Reading: Synonyms/antonyms Reading: Passage comprehension
0.701 0.586 0.606 0.560
* p < .01.
(0.042) (0.057) (0.028) (0.028)
z-score
Beta (ST)
p
6.303 2.461 3.783
0.691 0.982 0.773
Contents lists available at ScienceDirect
Research in Developmental Disabilities
The relationship between self-determination and academic achievement for adolescents with intellectual disabilities Amy S. Gaumer Erickson *, Patricia M. Noonan 1, Chunmei Zheng, Jennifer A. Brussow 2 University of Kansas, Center for Research on Learning, 521 JR Pearson Hall, 1122 West Campus Road, Lawrence, KS 66045, United States
A R T I C L E I N F O
A B S T R A C T
Article history: Received 2 May 2013 Received in revised form 8 September 2014 Accepted 9 September 2014 Available online 11 October 2014
Previous research has demonstrated that for students with intellectual disabilities, improved self-determination skills are positively correlated with productivity and organization during school and quality of life outcomes in adulthood. Despite extensive investigation in these areas, the predictive relationship between self-determination and academic achievement for students with intellectual disabilities has not been fully established. This study utilized the sample from the National Longitudinal Transition Study-2 of 480 adolescents with intellectual disabilities in the United States in an attempt to provide a possible empirical explanation of the relationship between academic achievement and self-determination, taking into account the covariates of gender, family income and urbanicity. The structural equation model was found to closely fit the data: all path coefficients were statistically significant. The results of this study identify a strong correlation between self-determination and academic achievement for adolescents with intellectual disabilities, indicating a linear relationship of these skills and supporting an increased focus on the teaching of self-determination skills. ß 2014 Elsevier Ltd. All rights reserved.
Keywords: Mental retardation Developmental disabilities Intellectual disabilities Self-determination Academic achievement
1. Introduction Academic achievement, typically defined as proficiency in reading and mathematics, has consistently been identified as a predictor of post-school success, including social inclusion, economic self-sufficiency, and overall quality of life (Day & Newburger, 2002; Greene, 2000; Kutner et al., 2007; O’Neill, 2001; Swanson, 2004). Moreover, research has explicitly related higher levels of achievement in reading, writing, math, and problem-solving skills with improved post-school outcomes for students both with and without disabilities (Benz, Yovanoff, & Doren, 1997; Schneider, Kirst, & Hess, 2003). In order to secure these positive post-school outcomes for students, educational policy and research continues to focus on promoting academic achievement. Scholarship in post-school outcomes for students with intellectual disabilities has identified self-determination as an important attribute. Specifically, researchers have documented (a) the efficacy of self-determination interventions for
* Corresponding author. Tel.: +1 785 864 0517. E-mail addresses: [email protected] (A.S. Gaumer Erickson), [email protected] (P.M. Noonan), [email protected] (C. Zheng), [email protected] (J.A. Brussow). 1 Tel.: +1 785 864 0593. 2 Tel.: +1 785 864 0517. http://dx.doi.org/10.1016/j.ridd.2014.09.008 0891-4222/ß 2014 Elsevier Ltd. All rights reserved.
46
A.S. Gaumer Erickson et al. / Research in Developmental Disabilities 36 (2015) 45–54
students with intellectual disabilities, (b) a positive relationship between self-determination and post-school outcomes for students with intellectual disabilities, and (c) a positive relationship between self-determination and access to the general education curriculum for students with intellectual disabilities (Agran, Blanchard, & Wehmeyer, 2000; Agran, Blanchard, Wehmeyer, & Hughes, 2001; Brooks, Todd, Tofflemoyer, & Horner, 2003; Craft, Alber, & Heward, 1998; Hughes et al., 2002; McCarl, Svobodny, & Beare, 1991; Moes, 1998; O’Reilly, Lancioni, Gardiner, Tiernan, & Lacy, 2002; Rock, 2005; Warner & de Jung, 1971). Furthermore, self-determination has been shown to be positively correlated with school engagement and adult outcomes for students with intellectual disabilities (Agran et al., 2005; Arndt, Konrad, & Test, 2006; Sa´nchez & Roda, 2003; Wehmeyer & Palmer, 2003; Wehmeyer & Schalock, 2001; Wehmeyer & Schwartz, 1997, 1998). Despite this body of research, studies overall exhibited a ‘‘lack of focus on academic skills in the self-determination literature for students with mental retardation/developmental delays’’ (Fowler, Konrad, Walker, Test, & Wood, 2007, p. 281). The current study incorporated structural equation modeling to investigate the direct relationship between self-determination and academic achievement for students with intellectual disabilities. 2. Theory Working within theories of human agentic behavior, Wehmeyer and Little (2009) describe a self-determined person as the ‘‘origin of his or her actions, [who] has high aspirations, perseveres in the face of obstacles, sees more and varied options for action, learns from failures, and overall, has a greater sense of well-being’’ (p. 868). Wehmeyer (2005) has also defined self-determined behavior as ‘‘volitional, intentional, and self-caused, or self-initiated action’’ (p. 115). Self-determined behavior differs from ‘‘other-determined behavior’’ by specifying self as the primary causal agent as opposed to behavior caused by external forces (Wehmeyer, Abery, et al., 2011). Self-determined behavior is affected by environmental factors as well as by individual knowledge, values, and skills (Field & Hoffman, 2001, 2007; Hoffman & Field, 2005). The influence of self-determination interventions on students with disabilities has been the focus of substantial research. Researchers have explored various methods to increase self-determination, including evidence-based practices that use choice-making, goal-setting, and self-advocacy to increase students’ self-determined behavior (Ezell, Klein, & Ezell-Powell, 1999; Fullerton & Coyne, 1999; Martin et al., 2003; Wehmeyer, Palmer, Lee, Williams-Diehm, & Shogren, 2011). Findings indicate that interventions are effective for improving self-determination (Chambers et al., 2007). Research has also examined the impact of self-determination skills on class participation and direction-following skills (Agran et al., 2005; Fowler et al., 2007; Konrad, Fowler, Walker, Test, & Wood, 2007; Lee, Wehmeyer, Soukup, & Palmer, 2010). Studies have used a variety of methods to assess the effect of self-determination interventions on quality and productivity in language arts and math; however, the existing research has focused primarily on skills such as organization and productivity rather than on academic achievement outcomes (Fowler et al., 2007). It is documented in the academic literature that individuals with developmental or intellectual disabilities are less likely to exhibit self-determined behavior due to their limited choice-making opportunities (Wehmeyer, Kelchner, & Richards, 1995, 1996; Wehmeyer & Metzler, 1995). As reported by Wehmeyer and Palmer (2003) and Wehmeyer and Schwartz (1997), higher self-determination scores were correlated with more positive post-school outcomes for students with intellectual disabilities. Additionally, Wehmeyer and Schwartz (1998) showed that self-determination skills corresponded with a positive quality of life for people with intellectual disabilities; this finding was replicated in an international survey of adults with mild intellectual disabilities (Lachapelle et al., 2005). Moreover, though it is a common assumption that cognitive limitations result in limitations in self-determination skills (Wehmeyer, Abery, et al., 2011), research has demonstrated that the opportunity to make choices was the most significant variable predicting levels of self-determination (Wehmeyer & Garner, 2003; Wehmeyer, Palmer, et al., 2011). Many moderating variables such as culture, gender, age, cognitive ability, religious beliefs, environment, and experiences of oppression have all been shown to impact self-determination (Duvdevany, Ben-Zur, & Ambar, 2002; Kurtz-Costes, Rowley, Harris-Britt, & Woods, 2008; Nota, Ferrari, Soresi, & Wehmeyer, 2007; Shogren et al., 2007; Wehmeyer, Abery, et al., 2011; Wehmeyer & Bolding, 1999, 2001; Wehmeyer, Palmer, et al., 2011). Student achievement is also influenced by moderating factors including socioeconomic status, gender, and urbanicity. For example, research has found that family and community poverty are frequently associated with negative effects on academic achievement (Blackorby, Chorost, Garza, & Guzman, 2004; Hattie, 2009; Sharkey, 2009; Wagner, Newman, Cameto, & Levine, 2006). The National Longitudinal Transition Study-2 (NLTS2) classifies additional covariates as demographic: age, ethnicity, family (e.g., support, expectations), and school characteristics and experiences (e.g., grade retention, absenteeism, behavior at school) (Wagner et al., 2006). The complex nature of the interrelationships between factors complicates research on the relationship between self-determination and academic achievement for students with intellectual disabilities. To advance scholarship on the relationship between self-determination and academic achievement, this study applied a structural equation model developed by the authors (Zheng, Gaumer Erickson, Kingston, & Noonan, 2014) that found a positive linear relationship between self-determination and academic achievement for adolescents with learning disabilities. To test the hypothesized linearity of the variables for adolescents with intellectual disabilities, two research questions were investigated: A. Is there a direct relationship between self-determination and academic achievement for adolescents with intellectual disabilities?
A.S. Gaumer Erickson et al. / Research in Developmental Disabilities 36 (2015) 45–54
47
B. Is self-determination a good predictor of academic achievement for adolescents with intellectual disabilities?
3. Methods 3.1. Participants This study is based on the data from Waves 1 and 2 of the National Longitudinal Transition Study-2 (NLTS2). NLTS2 was commissioned by the U.S. Department of Education, Office of Special Education Program (OSEP) in order to catalog the home and school experiences, academic performance, and post-school outcomes of a national sample of students with disabilities over a period of 10 years. This longitudinal data collection began in 2002 with Wave 1, which collected data on 11,000 youth across the United States aged 13–16, and concluded with Wave 5, which followed up with these same individuals at ages 23–26. Wave 1 and Wave 2 included surveys of parents, youth, and school staff as well as assessments of reading and mathematics achievement, self-determination, self-concept, and friendship. The direct assessments within these waves, which constitute the dataset for the current study, were conducted in 2002 and 2004 with youth 16–18 years of age. NLTS2 data mirrored the 12 disability categories (i.e., learning disabilities, speech impairment, mental retardation, emotional disturbance, hearing impairment, visual impairment, orthopedic impairment, other health impairment, autism, traumatic brain injury, multiple disabilities, and deaf/blindness) outlined in the Individuals With Disabilities Education Improvement Act (IDEA, 2004). The current study focused exclusively on participants identified with a primary disability of mental retardation reported on the students’ individualized education programs (IEPs), yielding a total sample size of 480 students. This sample is referred to as students with intellectual disabilities within the current study. The focus on a single disability category was purposeful, making these results more specific and descriptive than a combination of disability categories would allow. In line with national population trends, the dataset of youth with intellectual disabilities included 52% males, 54.8% white, 33.3% African American and 9.6% Hispanic students. The youth with intellectual disabilities represented individuals in low-income families (54.9%), single-parent households (34.5%), households that did not use primarily English at home (6.3%), and households with a head of household who did not complete high school (32.3%) (Wagner et al., 2003). 3.2. Measures & administration procedures The NLTS2 direct assessment data collection spanned six direct assessment domains: reading, math, content (i.e., science and social studies), self-concept, self-determination, and friendship interaction. The analyses within the current study focused on three of the six domains (reading, math and self-determination). These assessments were administered by trained personnel, typically school psychologists or teachers, who were hired and supervised by NLTS2 staff. Assessments were generally conducted at the adolescent’s school during non-instructional time. All instrument protocols were ‘‘fully scripted to maintain consistency of administration across assessors’’ (Wagner et al., 2006, p. 11). NLTS2 booklets were used to record the responses of each student and then sent to NLTS2 program staff for scoring (Wagner et al., 2006). 3.2.1. Academic achievement measure The NLTS2 assessors administered the Woodcock-Johnson Research Edition (Woodcock, McGrew, & Mather, 2007) to measure students’ academic achievement in math and reading. This research edition is a shortened form of the published Woodcock-Johnson III; its average reliability is .65 (Wagner et al., 2006). Though the published version contains additional items, both versions measure the same constructs and report on the same score scales. Two indicators for math (i.e., applied problems and calculation) and reading (i.e., synonyms/antonyms and passage comprehension), were calculated by NLTS2 staff (Wagner et al., 2006). Reported scores for math and reading were converted to z-scores within the current study. 3.2.2. Self-determination measure The Arc Self-Determination Scale measures individual differences in students’ tendencies to function in a self-determined way (Wehmeyer, 1995). The 72-item measure, with items categorized into one of four subdomains (i.e., autonomy,self-regulation, self-realization, and psychological empowerment) was normed on a sample of 500 youth, primarily identified as having intellectual and learning disabilities, reporting a coefficient alpha of .90. The NLTS2 researchers modified the Arc Self-Determination Scale by reducing the number of items in each domain and deleting the subdomain of self-regulation (Facts from OSEP’s National Longitudinal Studies, 2005). The modified scale contained 26 items across three subdomains: autonomy (15 items), self-realization (five items), and empowerment (six items). Autonomy was measured by a four-point scale: (1) not when I have the chance, (2) sometimes, (3) most of the time, and (4) every time I have the chance. A four-point scale was also used for self-realization questions: (1) never agree; (2) sometimes agree; (3) usually agree; and (4) always agree. Empowerment items utilized a two-point scale. Of these items, one represented passive and pessimistic actions (e.g., ‘‘It is no use to keep trying because it will not change things’’) and the second represented active and optimistic actions (e.g., ‘‘I keep trying even after I get something wrong’’).
48
A.S. Gaumer Erickson et al. / Research in Developmental Disabilities 36 (2015) 45–54
3.3. Analysis The structural equation model incorporated into the data analysis was replicated from a model that utilized a NLTS2 sample of students with learning disabilities (Zheng et al., 2014). The results of the previous study showed a significant positive relationship between self-determination and academic achievement for students with learning disabilities. To test the validity of the model, the same structural equation methodology was applied to the NLTS2 sample of students with mental retardation as their primary IDEA disability classification (referred to as students with intellectual disabilities within the current study). Because the scales varied within the assessments, all scores were converted to z-scores for analysis. Research questions were tested via structural equation modeling following four steps: (1) first-order confirmatory factor analysis (CFA) to specify the relationship between observed and latent variables and the relationships among the latent variables; (2) second-order CFA with beta coefficients to test the significance of these relationships; (3) latent regression between self-determination and academic achievement to test the linearity of the relationship; and (4) inclusion of the covariates gender, family income, and urbanicity in the beta coefficients to test the significance of the overall model. The first-order CFA included five variables: math (M), reading (R), autonomy (AU), self-realization (SR), and empowerment (EM). In the second-order CFA, academic achievement (AA) included the math and reading as subdomains, and self-determination (SD) included autonomy, self-realization, and empowerment as subdomains. The covariates were coded into categories in this data set. ‘‘Male’’ served as the baseline for the gender comparison (Gen); thus, Gender 1 (Gen1) was the female group. Family incomes below $25,000 constituted the baseline group for income (Inc); Income 1 (Inc1) referred to family incomes between $25,000 and $50,000, and Income 2 (Inc2) referred to family incomes over $50,000. Youth from rural areas were the baseline group for urbanicity (Urb), which resulted in Urbanicity 1 (Urb1) referring to suburban areas and Urbanicity 2 (Urb2) referring to urban areas. The three variables of autonomy, self-realization, and empowerment were parceled to form manifest indicators for each of the latent variables. A parcel is defined as ‘‘an aggregate-level indicator comprised of the sum or average of two or more items, responses, or behaviors’’ (Little, Cunningham, Shahar, & Widaman, 2002, p. 152). The advantages of parceling over item-level modeling include higher reliability, small and equal intervals, and a decreased chance of distributional violations. An item-to-variable balancing technique was used to create the parcels for the three variables. Item-total correlations were computed using SPSS, and items of higher correlations were then combined with items of lower correlations. For the five items in the self-realization domain, items 1 and 5 were parceled as the first indicator, items 2 and 4 as the second indicator, and item 3 was the third indicator. Since the scales of measurement varied widely across variables, each indicator was standardized across all variables to have a mean of 0 and a standard deviation of 1.0. The variance was specified as 1.0 for both the first- and second-order variables in order to establish common scales for all variables. This process allowed the variance/covariance matrix of all variables to produce a correlation matrix. Factor loading equations were used to improve the model identification because there were only two items for both the math and reading indicators. 3.4. Missing data The data set had a very small proportion of missing data, on average 3.5% for each variable. In an effort to improve the accuracy of statistical analyses, the multiple-imputation method was performed to impute the missing data and obtain a complete data set. The proportion of missing data was small enough that only one repetition of the multiple imputation method was necessary. All variables (e.g., math score, reading score, gender, and family income) were considered for the imputation in order to minimize bias and increase the accuracy of parameter estimates. 3.5. Weights The stratified and clustered sampling design of the NLTS2 provided a robust dataset; however, this design incorporated proportional oversampling and undersampling of some population subgroups, thus both sampling and replicative weights were provided in the original dataset. Because clustered sampling can result in case similarities, replicative weights were included in data analyses. These replicative weights were intended to reduce the estimate bias and account for the independence of observed cases. Additionally, sampling weights ensured that analyses accurately represented the population of students with intellectual disabilities in the United States. 4. Results Sample sizes have been rounded to the nearest 10 for all results as per the IES restricted-use data requirements of disclosure protection (Institute for Education Sciences, 2012). For the 480 youth with intellectual disabilities, the average math and reading z-scores were 0.74 (SD = 0.81) and 0.86 (SD = 0.78), respectively. The 420 respondents’ average ratings on each of the self-determination subscales were skewed toward positive options. For example, the two-point empowerment scale had a mean of 1.80 (SD = 0.19) indicating that most students viewed themselves as more empowered than not. On the four-point self-realization scale, the mean of 3.08 (SD = 0.55), indicates that most students considered themselves to be highly self-regulated. The four-point scale for the autonomy domain yielded a mean of 2.93 (SD = 0.47), suggesting that most students believed they typically acted autonomously.
A.S. Gaumer Erickson et al. / Research in Developmental Disabilities 36 (2015) 45–54
49
Table 1 Correlations among the academic achievement and self-determination variables for adolescents with intellectual disabilities. Construct
Academic achievement
Autonomy
Self-realization
Empowerment
Academic achievement Autonomy Self-realization Empowerment
1.000 0.208 0.286 0.553
1.000 0.738 0.505
1.000 0.627
1.000
4.1. Factor analysis The factor analysis identified positive correlations between academic achievement and each of the self-determination subscales (see Table 1). The three self-determination subscales demonstrated strong correlations, indicating a clear pattern among the first-order variables. Furthermore, each self-determination subdomain was found to correlate positively with academic achievement (i.e., combined reading and mathematics proficiency). These correlations imply that students with higher levels of autonomy, self-realization, and/or empowerment exhibit higher levels of academic achievement. All factor loadings were significant. The reading, mathematics, and each of the three self-determination subscales all had factor loadings significant at the .01 level, and each variable’s z-score was above the critical value of 1.96. The standardized Beta results represent each variable’s standardized factor loadings. After controlling for gender, family income, and urbanicity, all standardized Beta statistics except for self-realization resulted in significant p values at the .01 confidence level (see Table 2). The self-realization subscale seemed to have been substantially influenced by the covariate variables, which may explain its exceptionality. Overall, factor loadings indicate that the variables of interest were strongly related. 4.2. Structural model To depict the relationship between academic achievement and self-determination after controlling for gender, family income, and urbanicity, a model with these variables as covariates was generated (see Fig. 1). The analysis found self-determination to be a significant predictor for academic achievement even after controlling for the covariant variables (b = 0.380, p = .001), as shown in Table 3. Urbanicity was also a significant predictor for academic achievement, with students attending urban schools having significantly lower levels of academic achievement. The results did not identify family income or gender as significant predictors of academic achievement for adolescents with intellectual disabilities. Although one of the three covariates was found to be a significant predictor of academic achievement, self-determination remained the strongest predictor of academic achievement even when controlling for gender, family income and urbanicity (beta = 0.350, p = .001). Interpretation of the standardized regression coefficient (Beta ST) reveals that an increase of one standard deviation in the level of self-determination results in an increase of .35 standard deviation for achievement in reading and mathematics for students with intellectual disabilities. This finding supports a direct relationship between self-determination and academic achievement for adolescents with intellectual disabilities. The structural equation model showing the second-order factor loadings and the overall standardized Beta statistics after controlling for gender, family income, and urbanicity can be found in Fig. 1. 4.3. Limitations This study’s primary limitation can be found in the measures themselves. NLTS2 only incorporated selected items from the self-determination scale (i.e., 26 of the 72 items in the Arc Self-Determination Scale). Additionally, an entire subscale, self-regulation, was not included in the NLTS2 data collection. Given this limitation, we cannot predict whether the linear relationship identified by this analysis would have been consistent if the full version of the Arc Self-Determination Scale had been administered. Table 2 Structural equation model: factor loadings after adding covariate variables. Construct
Factor loadings (SE)
Self-determination Autonomy Self-realization Empowerment
0.956 (0.152) 5.205 (2.115) 1.218 (0.322)
Academic achievement Math: Applied problems Math: Calculations Reading: Synonyms/antonyms Reading: Passage comprehension
0.701 0.586 0.606 0.560
* p < .01.
(0.042) (0.057) (0.028) (0.028)
z-score
Beta (ST)
p
6.303 2.461 3.783
0.691 0.982 0.773
