Author: Stewart, James F
Date published: April 1, 2010
(ProQuest: ... denotes formulae omitted.)
In today's learning environment, technology has become enmeshed within the educational landscape of many college campuses. The growing availability of inexpensive high speed Internet access, as well as multimedia academic and productivity technology (i.e., Blackboard, Microsoft PowerPoint, Smartboards) as tools for face-to-face (traditional), online, and hybrid classes, has been a boon for distance learning programs (Buckley & Smith, 2007; Hanna, 1998; Mbilinyi, 2006). In fact, 3 1 .5% of higher educational institutions in the U.S. offered online courses in 2005, while 3.18 million students took at least one online course during the same year. The latter represent an increase of approximately 60% from two years prior (Allen & Seaman, 2006).
Online learners are often described as more nontraditional students who may be older and/or have disabilities (Allen & Seaman, 2006; Carr, 2000; Githens, 2007). Improvements in technology have created greater opportunities for learning among those with disabilities by providing broader access to education, information, and learning (Li & Irby, 2008). However, technology is not perfect, and many students with disabilities (SWD) continue to have trouble accessing some of the technology employed in the classroom (Burgstahler, 2002). If a webpage uses an incompatible programming language, fails to attach descriptions to pictures, or utilizes instruction tools such as Microsoft PowerPoint (when embedded as a raw .ppt or .pps file), using assistive technology such as a screen reader for the Blind may be futile (Edmonds, 2004; Georgia Tech Research on Accessible Distance Education Project, n.d.; Hart, 2008).
Educational Outcomes for Students with Disabilities
Opportunities to participate in postsecondary education have significantly increased for students with disabilities. Legislation, including Title I and III of the Americans with Disabilities Act of 1990 (ADA) and Sections 504 and 508 of the Vocational Rehabilitation Act of 1973 (Rehab Act), have provided legal support for students with disabilities in the classroom. Students with disabilities are matriculating to college in increasing numbers, represented by a 17% increase between 1987 and 2003 (Wagner, Newman, Cameto, & Levine, 2005). Furthermore, students with disabilities represent anywhere between 9% and 15% of the college students in the U.S.. (Frieden, 2003; Harbour, 2004; HEATH, n.d.; Horn, & Nevill, 2006; Kinash, Crichton, & KimRupnow, 2004; Walker, Turner, Haile-Michael, Vincent, & Miles, 1995; Weathers et al., 2007).
Despite increased postsecondary attendance, a large majority of the existing literature purports that SWD do not score as well, attend, or graduate from four-year colleges and universities at levels which are comparable to peers without disabilities (Murray & Wren, 2003; Tincani, 2003). In a longitudinal study, Murray, Goldstein, Nourse, and Edgar (2000) explored the postsecondary academic outcomes of students diagnosed with learning disabilities (LD). Two separate dataseis spanning the years from 19851990 and 1991-1995 were utilized to determine whether any cohort differences existed across the 10 year period (passage of ADA possibly benefitting SWD more so in the latter cohort). The purpose of the study was to determine whether attendance and graduation rates of individuals with and without disabilities differ in postsecondary educational settings. The researchers collected self-reported data from a sample of 1 68 SWD and 3 1 5 students without disabilities (SWOD) who graduated from three different K- 12 school districts. Using chi-square analyses, the findings purport that during each of the 10 years after graduation, secondary school graduates with LD were significantly less likely to attend college when compared to peers without a disability (_?<.001).. Similarly, men with disabilities yielded lower rates of attendance when compared to counterparts without (p<.05). The same was true when comparing the two groups of women. Despite using descriptive statistics rather than inferential to explore the rate of graduation between the groups, there is a clear difference between them. At both 5 and 10 years post-graduation, only 1 .2% and 2.4% SWD (respectively) had completed college. In contrast, 24.2% and 45.5% of SWOD obtained a four-year degree during the same respective periods. Surprisingly, no women with disabilities in this study obtained a four-year degree 10 years after high school, unlike other literature purporting similar or greater rates of college attendance among women in the general population (Diaz, 2000; Horn & Nevill, 2006; Planty, Hussar, Snyder, Provasnik, Kena, Dinkes, et al., 2008). However, this may be attributable to the small number of women in each of the LD cohorts.
Similarly, Wagner et al. (2005) utilized longitudinal data to investigate outcomes for secondary school students with disabilities. The researchers analyzed dataseis from the National Longitudinal Transition Study (NTLS) and the National Longitudinal Transition Study-2 (NLTS-2) to explore changes in educational outcomes between 1987-2003. The NTLS included data from a sample of secondary-school-age students who were receiving special education services in 1985 and corresponding student/parent reports from 1987 (Cohort 1). The NTLS-2 consisted of student/parent reported data for all students receiving special education services in 2003 (Cohort 2). The purpose of the research was to ascertain what changes students from Cohort 1 and Cohort 2 experienced between 1987 and 2003. Students between the ages of 15 and 18 years experienced significant increases in graduation rates from 1987-2003. Additionally, students in 2003 who were two years (or less) removed from the ?12 system exhibited a significant increase in four-year college attendance (p<.01). Despite positive strides, the researchers contend that a significant discrepancy still exists between SWD and SWOD for each of the aforementioned outcomes.
While the aforementioned research focused primarily on disability status and other demographics as independent variables for educational outcomes, Trainin and Swanson (2005) expand on this research by including cognitive constructs as independent variables. Specifically, the researchers investigated, while controlling for age, whether cognitive ability and metacognition (i.e., motivational and learning strategies) of students with and without disabilities were related to academic achievement. The researchers found that SWD exhibited significant deficits in basic reading processes (i.e., word reading, pseudo-word reading, & real word reading), semantic processing tasks, and processing speed. Nevertheless, SWD exhibited superior (p < .05 for both) help seeking behaviors (e.g., asking for assistance from friends and teachers) and self-regulation of learning and time management (e.g., studying longer or early), which the researchers suggest act as compensatory strategies that may attribute to similar level of achievement between the two groups.
Similarly, Seo, Abbott, and Hawkins (2008) conducted a longitudinal analysis of postsecondary outcomes for 60 SWD and 5 1 1 SWOD, employing multinomial logistic regression analyses while controlling for gender, ethnicity, and SES. Using "no postsecondary school attendance" as a reference revealed comparable rates of full-time and part-time college attendance among both groups at age 24. Yet, there was a significant difference between full-time college participation at age 2 1 . In contrast, college graduation rates of SWD and SWOD did not significantly waiver at either 21 or 24 years (while using high school graduation as a reference point).
The existing literature exploring the academic outcomes for those with disabilities illustrate contradictory representations of grades, college participation, and graduation rates. Seemingly objective data, such as census and large survey records, give the impression that SWD are considerably worse off academically than SWOD (Murray et al., 2000; Wagner et al., 2005). Conversely, contemporary research that accounts for extraneous variables, such as grades, age, and gender, reveals comparable academic outcomes for SWD and SWOD (Klinkosz, Sekowski, & Brambring, 2006; Seo et al., 2008; Trainin & Swanson, 2005). Our research will expand upon this current research, while recognizing multiple level effects that impact grades.
Online vs. Traditional Courses
Historically, online learning environments have been viewed as inferior alternatives to traditional "bricks-and-mortar" courses. Both college educators and administrators have questioned whether the dynamic environment that exists in the classroom can be replicated online (Aragon, Shaik, Palma-Rivas, & Johnson, 1999). Nevertheless, perception of the quality of online education has improved among both college personnel and students in recent years (Allen & Seaman, 2006). Despite existing concerns about the efficacy of online courses, a lack of quantitative empirical studies exists that explore how comparable online and traditional courses really are. Much of the research literature which explores the efficacy of online learning programs has been descriptive in nature and not theoretically robust (Lee, Driscoll, & Nelson, 2004).
To fill this void, Aragon et al. (1999) implemented a causal comparative study to explore student outcomes of two convenience samples of 19 graduate students registered for one online and one traditional course (identical courses), each taught by a single instructor. The researchers explored whether level of student satisfaction (i.e., instructor effectiveness and overall course quality) and student learning outcomes (i.e., course grade, and self-efficacy on course content) varied among classes. Separate independent samples t-tests were used for each of the analyses. Traditional students (M=4.2l, SD=.79) were slightly more satisfied with the instructor effectiveness than online students (M=3.58, SD=I. 07), t(36) = 2.07, p < .05, while there was no difference between overall course satisfaction of the two groups. Self-efficacy did not differ across groups in approximately 83% of the content areas regarding ability to implement course content in the field. Of the five learning areas that students differed in level of self-efficacy, online students only felt more prepared in one. Despite the aforementioned dissimilarities, there was no significant difference between the grades of students taking online or traditional courses
Bertsch, Callas, Rubin, Caputo, and Ricci (2006) conducted a quasi-experimental study to explore whether student achievement differs for those who take courses using a distance learning or traditional learning model. Utilizing a crossover design, the researchers compared a convenience nonrandom sample of 52 medical students who were separated into two equal groups. Each group received two months of weekly lectures by means of a traditional and vidéoconférence format. One group received one full month of lectures using one format, while the other simultaneously received the alternative format. Course format was reversed for each group during the second month of the course. At the outset of the two month course, each participant was assessed to determine his or her ability to learn the course content by completing the Clinical Practice Examination (CPX), which included the subject matter conveyed during the two month lectures. Using Analysis of Variance (ANOVA), the researchers found that there was no significant difference between the percentage of students who successfully completed the CPX questions for content covered using either a traditional format (M=I 6, SD=M) or distance learning format (M=78, SD= 12). This insignificant finding supports the notion that students were equally successful with learning the course content using either format (p>?5).
Demographics such as undergraduate cumulative GPA, gender, and course level (undergraduate or graduate) have been established in the literature as significant covariates of postsecondary academic success for students with and without disabilities (Aragon, Johnson, & Shaik, 2002; Chyung, 2007; Halberstam & Redstone, 2005; Retsas & Wilson, 1996). However, learning does not happen in a vacuum, and elements that influence learning are often interconnected. The diversity of online communities warrants the need to explore what effect covariates (e.g., cumulative GPA, gender, and credit hours earned) have on student grades across multiple levels. Likewise, these relationships using a nested structure of data (i.e., when students are nested within class) must be investigated. The collective impact of the aforementioned variables varies from teacher to teacher and class to class, which in turn can affect statistical analyses, and interpretations that follow. Yet, the need for using nested structures is not prevalent in the current literature pertaining to academic outcomes for students with disabilities who take online courses. Utilization of Academic Search Premier, ERIC, and ProQuest academic search databases yielded no empirical studies, investigating online educational outcomes for SWD and SWOD, while controlling for covariates in a nested model. The dearth in the coverage of this important topic and the need to fill this void in the literature is the impetus for this study.
The current study uses data from a purposive non-random sample of students taking online and traditional face-to-face courses during the Fall 2006 and Spring 2007 semesters to answer the following four research questions:
1. Does disability status have an effect on student outcome (course grade) after controlling studentlevel covariates (cumulative GPA, gender, and total credit hours accumulated), while accounting for classroom differences?
2. Do disability status effects on student outcomes (class grade) vary over classrooms after controlling for student-level variables (cumulative GPA, gender, and total credit hours accumulated)?
3. Does course delivery format (online or offline) have an effect on student outcomes after controlling class room level covariates (type of course, and level of course)?
4. Does course delivery format (online or offline) have an effect on the effect of disability status on class grades, after controlling classroom level covariates (type of course, and level of course)?
The study sample was comprised of full-time and part-time students at a midsized, Historically Black College or University (HBCU) on the East Coast of the United States. The researcher employed the use of a purposive non-random sample of students (n=3,078) from the Fall 2006 and Spring 2007 semesters to answer each of the research questions set forth in the study. The sample included both undergraduate and graduate students with (n=157) and without disabilities (n=2921), the majority of whom were African American (>95%). Members of the sample designated as having a disability included all students who registered with the institution's Disability Support (DS) program and were receiving reasonable accommodations, as mandated under ADA and Section 504 of the Rehab Act. Qualification for accommodations was established according to the institution's disability documentation requirements, which are based on guidelines recommended by the Association on Higher Education and Disability (1997) and the Educational Testing Service (1999, 2001, 2003, & 2007). The disability types represented within the sample included 20% LD, 20% Psychiatric, 16% Visual/Perceptual, 12% Neurological, 10% ADD/ADHD, and 10% Multiple Disabilities. Sample members without disabilities included all students (not enrolled with DS office) who participated in any course taken by a student who qualified for a reasonable accommodation.
The procedure used while collecting data for the current study was completed using a dual step process. At the outset, data were obtained from the University Registrar's office, for all classes taken during the 2006-2007 academic year by students who qualified for reasonable accommodations due to disability. This data consisted of student ID number for all students in each course (replaced with independent numeric code after dataset was coded and organized); cumulative GPA; cumulative career credit hours earned; gender; class type (major/elective or general education course); academic level (undergraduate or graduate student); and final grade submitted for course. To code for class level nesting affects, the following data were also collected from the Registrar: instructor name; course code; course title; number and section; and semester (i.e., Fall/Spring). Each of these five variables was eliminated from the dataset after the 143 courses included as part of the study were provided independent numeric codes.
The second step of the data collection process consisted of eliminating variables from the dataset that might bias the results. Any student who did not receive a grade of an "A", "B", "C", "D", or "F", was excluded from the sample. In other words, students who received a withdrawal (W), continuing study (CS), or pass/fail (P/F), as a grade met the criteria for exclusion. Consequently, 22 courses were eliminated from the dataset because the only SWD in the class met the criteria for exclusion. Removing each of these courses eliminated the possibility of biased nesting affects based on courses with no SWD receiving a grade.
To address each of the research questions, the outcome variable grade (the grade for the current class) is represented as consecutive integers from 0 to 4 (O=F, I=D, 2=C, 3=B, 4=A). Additionally, the following student-level independent variables will be examined: disability status (X^sub 1^; disstat), cumulative GPA (X^sub 2^; cumgpa), gender (X^sub 3^; gender), and the number of hours completed at the university (X^sub 4^; credhrs). X^sub 1^ (disstat) is a dichotomous variable dummy coded for students without disabilities as zero and students with disabilities as unit. X^sub 2^ (cumgpd) is a metrical variable calculated as a mean score of all grades from classes the students have taken thus far. X^sub 3^ (gender) is a dichotomous variable dummy coded male as zero and female as unit. X^sub 4^ (credhrs) is a metrical variable which represents the cumulative total of course hours taken thus far.
The class-level variables included whether a class is on-line or traditional format (Wx; mode). For the purpose of this study, traditional courses were considered as those that occurred in a traditional classroom setting (or comparable), without any significant online component integrated into course instruction. While many courses utilize online components (e.g., posting syllabi, online submission of assignments and quizzes), only courses that were used to facilitate learning through internet based instruction (e.g., online lectures, student/instructor discussions) at least 50% of the class were interpreted as online. The second class-level covariate was classified as either a major/elective course or general education (W^sub 2^; type), The demarcation of general education and major/elective courses recognizes the former as a course imposed onto all undergraduate students, regardless of major, or interest in the subject matter. A major/elective course is one that a student has chosen freely because it is an elective, or is part of his or her chosen field of study (and consequently, may be of greater interest to the participant). The nature of the final class-level variable, whether the course was an undergraduate or graduate course (W^sub 2^; level), is self explanatory. The class characteristics have dichotomous dummy codes for type (major/elective=0, general education=1), mode (traditional=0, online= 1), and level (undergraduate=0, graduate=l). Descriptive statistics for both student and class level variables are illustrated in Table 1 .
Hierarchical Linear Modeling
Hierarchical linear modeling (HLM; Raudenbush & Bryk, 2002) was used in the current study to analyze multilevel (both student and class level) covariates' effects on the outcome. HLM has several advantages over single level analysis techniques, such as ANOVA and Multiple Regression. The assumption of independence of cases is not necessary in a multilevel analysis because the probable dependence of students in the same classroom is explored explicitly by counting the nested structure of data (Aitkin & Longford, 1986). Also, aggregation bias is avoided because the HLM technique allows investigation of the effects without aggregating data because this multilevel technique allows analyses to be simultaneously conducted at multiple levels of data. In the current study, both student- and classroom-level sources of variability in outcomes are simultaneously accounted for by specifying a two-level HLM. Student demographic variables (e.g., gender, cumulative GPA) were modeled at the individual level and fixed at the classroom level, treating them as covariates to the class effects. Classroom level independent variables/covariates (e.g., class modality, class level, class type) were entered to help reduce the unexplained variance attributed to the outcomes. This technique helps examine the direct effects of the key level- 1 independent variable (disability status) as well as the covariates (e.g., gender, cumulative GPA), modeling both at the individual and classroom level simultaneously on the outcome variable. The program Hierarchical Linear and Nonlinear Modeling (HLM), version 6.0.1 (Raudenbush, Bryk, Cheong, & Congdon, 2004) was used for data analyses.
Figure 1 shows the two-level structure of our model. In this study, we have attempted to measure the effect of individual factors on grade, the effects of classroom factors on grade, and the effect of classroom factors on the relationship between disability status and grade.
HLM analyses had three stages for this study. The first stage was a Random Coefficient Model (Raudenbush, & Bryk, 2002). This model includes the student-level predictor variables only while counting classroom differences. First, this model is used to estimate how much variation in the outcome variable (grade) is attributed to individual level characteristics. This model also allows us to estimate the variability in the level- 1 intercepts and slopes over level-2 units (classrooms).
The second stage of HLM employed is referred to here as the Intercept as Outcome Model, as indicated by Raudenbush and Bryk (2002). This model includes classroom-level characteristics and student-level characteristics to explain the variability of class averages.
The final model is referred to as the Intercept and Slope as Outcome Model, and includes class-level and student-level characteristics. This model uses student-level and class-level characteristics to explain not only the variability in intercepts, but also the difference in the effect of disability status on grades among classes.
Random Coefficient Model
At the outset, all student-level independent variables (cumGPA, gender, and credhrs) were added to the level- 1 model to address the first research question. Additionally, we let the intercept (classroom average) and the effect of disability status on grade vary over classrooms. In other words, we treat intercept (β^sub 0^) and the slope (β^sub 1^) as random coefficients. This particular type of HLM model is defined by the following equations:
Level 1 Model: Y^sub ij^ = β^sub 0j^ + β^sub 1j disstat + β^sub 2j^; cumgpa + β^sub 3j^: gender + β^sub 4j^ credhrs + r^sub ij^
Level 2 Model: β^sub 0j^ = γ^sub 00^ +u^sub 0j^
β^sub 1j^ = γ^sub 10^+u^sub 1j^
β^sub 2j^ = γ^sub 20^
β^sub 3j^ = γ^sub 30^
β^sub 4j^ = β^sub 40^
where Y^sub ij^ is the course grade of ith student jth class, r^sub ij^ is ith student jth class residual, u^sub 0j^ jth class residual of β^sub 0^, and u^sub 1j^ is jth class residual of β^sub 1^,
The information in Table 2 illustrate the results of the hypothesis test addressing whether disability status affects student outcomes (course grade) after controlling student-level covariates (cumulative GPA, gender, and total credit hours accumulated), while accounting for classroom differences. In addition, associated null hypotheses are:
H0^10 = O
H^sub 0^: γ^sub 10^ = °
H^sub 0^: γ^sub 20^ = 0
H^sub 0^: γ^sub 40^ = 0.
Based on the test statistics, we found the latter three null hypotheses can be rejected, and credhrs, cumgpa, and gender are all statistically significant covariates for grade. However, disability status is not statistically significant (p = 0.782). In Table 2 (and subsequent tables), the Fixed Effects column specifies each fixed (not varying over subject or class) effect and the Coefficient column contains the estimates of these fixed effects, which can be interpreted as regression intercept or slope coefficients. In contrast, the Random Effects column specifies each random (varying over subject or class) effect and the Variance column contains the estimates of these random effects which are variance estimates. Thus, these values imply the quantities of unexplained variability (i.e., residual variance).
When examining the test statistics on the variance (random effects) part, we can infer that statistically significant differences exist in average grade among classes. Thus, we can reject the null hypothesis, H^sub 0^: τ^sub 00^= 0. The estimated variance of the slopes is examined to determine whether the null hypothesis, H^sub 0^: τ^sub 11^= 0 can be rejected. The test statistic associated with this hypothesis is not significant (p = 0.211); therefore, we cannot reject the null. Accordingly, we infer that the regression coefficient of disstat (ß^sub 1^) does not statistically significantly vary among classes.
The intra-class reliability statistics associated with ß^sub 0^ and ß^sub 1^, are (ß^sub 0^) = 0.821 and (ß^sub 1^) = 0.080 respectively. These estimates show that the intercepts are quite reliable (i.e., classroom difference explains a large amount of variability of classroom grade averages), while the slope estimates are far less reliable (i.e., classroom difference explains a very small amount of variability of classroom grade averages).
Intercept as Outcome Model
In addition to the finding from the first (random coefficient) model that the intercept term was varying among classes significantly, the second (intercept as outcome) model tests whether the intercept variance is significantly explained by the class-level covariates. While the student-level model remains the same, the class-level model is now expanded to include mode (online or traditional course), type (major/elective or general education), and level (undergraduate or graduate) to attempt to explain the variability in average grade across classes. This HLM model is defined by the following equations:
Level 1 Model: Y^sub ij^ = ß^sub 0j^ + ß^sub 1j^disstat + ß^sub 2j^ cumgpa + ß^sub 3j^gender + ß^sub 4j^credhrs + r^sub ij^
The hypothesis test results in Table 3 show that the class-level characteristics type (major/elective or general education) and level (undergraduate or graduate) did not significantly explain the variability in average grade across classes. In other words, the null hypotheses H^sub 0^: γ^sub 02^ = o and H^sub 0^: γ^sub 03^ = 0 were retained with alpha level 0.05. However, the null hypothesis, H^sub 0^: γ^sub 01^ = 0 was rejected with alpha level 0.01. In other words, traditional courses yielded statistically significantly higher mean grades than online courses when controlling for the effects of level and type.
When examining the Variance-Covariance components for the null hypothesis H^sub 0^: τ^sub 00^= 0, the associated x^sup 2^ value was 930.398 with df= 137 and p < .01. Unexplained variation in the intercepts remains statistically significant even after controlling for the class-level variables.
Intercept and Slope as Outcome Model
The third model includes student-level and class-level characteristics to attempt to explain the variability of both intercept (ß^sub 0^) and the random regression slope coefficient (ß^sub 1^). This HLM model is defined by the following equations:
Level 1 Model: Y^sub ij^ = ß^sub 0j^ + ß^sub 1j^ disstat + ß^sub 2j^ cumgpa + ß^sub 3j^ gender + ß^sub 4j^ credhrs + r^sub ij^
As seen in the level 2 equations, we treat not only the intercept (ß^sub 0^), but also the regression slope of disability status (ß^sub 1^) as outcome variables with four class-level variables as independent variables. In this model, we are mostly interested in whether the class format (mode) is a statistically significant independent variable after controlling two other class level covariates (type and level) by testing the null hypothesis, H^sub 0^: γ^sub 11^ = 0.
Note that this parameter of interest, γ^sub 11^, is also called crosslevel interaction term because this model parameter eventually becomes the parameter of the product term between the first-level independent variable (disstat) and the second-level independent variable (mode).
In Table 4, the results show parameter γ^sub 11^ to be statistically significant, with an alpha level of 0.05. This result tells us that there is a statistically significant tendency for students with disabilities to have higher grades in online courses versus traditional, after controlling the type and level of class. This is a particularly interesting observation given that disability status as a whole was not statistically significant according to the earlier analyses in this study. Yet, the cross-level interaction between disability status and course delivery format is significant. We will further discuss this interesting finding in the discussion section of this paper.
The purpose of the current study was to explore whether there is any significant difference between the outcomes of students with and without disabilities, who take online and/or traditional courses. The sample included both undergraduate and graduate students from a midsized HBCU on the East Coast of the United States. Accordingly, the sample was comprised mainly of African Americans.
We used two-level hierarchical linear modeling to investigate what effect various student-level (i.e., disability status, cumulative GPA, earned credit hours, gender) and class-level covariates (i.e., online/traditional, required/non-required, graduate/undergraduate) might have on college course grades. The first model, a random coefficient model, was used to explore whether the effect of disability status on grades varies over classrooms. This model established that: 1) on average, SWD did not score any different from SWOD when controlling for all other student level covariates, and 2) the disability status effect does not significantly differ among classes. The congruity of the students' grades is similar to other recent studies, which counter the historical belief of academic inferiority among SWD by accounting for other covariates (Murray et al., 2000; Murray & Wren, 2003; Tincani, 2003; Wagner et al., 2005).
In contrast to the insignificant differences between SWD and SWOD, gender, cumulative GPA, and credit hours earned, did have a significant effect on average grades within classes. Given that cumulative GPA and earned credit hours are likely to be significant correlates with higher course grades (Morris, Wu, & Finnegan, 2005), the finding was predictable. Current literature has recognized an increased level of academic performance among women in college, whose graduation rates now outpace those of men (Chyung, 2007; Diaz, 2000; Kim & Conrad, 2006; Marklein, 2005). This trend is magnified among African American men between the ages of 20-24, whose rate of graduation is outnumbered by female counterparts by more than 40% (Snyder, Dillow, & Hoffman, 2008). Consequently, our finding that women typically obtain higher grades than men was expected (as the sample was largely African American). The current research improves upon existing disability research by accounting for the nested structure of data using HLM. Our results show that disability status did not have any effect on grades across classes. It is important to recognize that neglecting to control for student level covariates, as well as nesting effects, could have produced quite different results (e.g., significant effect of disability status on grades).
The intercept as outcomes model was used to determine what class level covariates contributed to the remaining variation in average grades across classes, while holding both student level and class level covariates constant. This analysis confirmed a significant main effect for mode of instruction, revealing that, on average, student grades in traditional courses were significantly higher than online courses. Our findings run contrary to assertions in the existing literature that grades for online and traditional courses are comparable (Bertsch et al., 2006; Aragon et al., 1999). One explanation for this is that students in traditional classes usually harbor more positive feelings about an instructor's teaching style and interactions/feedback, compared to distance learning classes (Aragon et al., 1999). Another explanation for this finding is that the difference in results could be attributable to methodological differences between other studies that did not employ HLM, which accounts for multilevel structure of data. In contrast to the significant finding for mode of instruction, no main effect was established for course type or level.
The most notable of our results was the cross-level interaction of the effect of course mode (online or offline) on the relationship between disability status and class grade. The third research model, Intercept and Slope as Outcome Model, indicates that SWD perform better in online courses after controlling for the student- level characteristics ?? cumgpa, gender, and credhrs. This finding supports the assertion that SWD may benefit significantly from modern, multimedia rich online course formats, which are characterized by the interchangeable use of text, audio, video, pictures, graphics, and animation (Buckley & Smith, 2007). Modern online teaching methodologies allow instructors to cater to a wide variety of learners with different learning styles, while finding new ways to engage student learning. The capability of utilizing a variety of instructional styles under the online teaching model allows students with different learning styles (e.g., visual and auditory learning, & internal processing) to learn just as well as those attending traditional courses (Aragon et al., 2002).
Growing competition within the U.S. workplace has necessitated continued learning to help ensure greater opportunities for employment (Maurer, Weiss, & Barbeite, 2003). The growing income deficit between those who graduate from high school and college clearly illustrates these changes (U.S. Dept. of Education, 2007). Results of the random coefficient model used in our study demonstrate that the grades of SWD are comparable to peers, which may inform vocational rehabilitation (VR) counselors who are providing services to those with disabilities. The distribution of federal and state funding to provide VR services are allocated to provide job development and training commensurate to one's ability and desire. Our findings may serve as a catalyst to inform VR counselors about the positive outcomes SWD can achieve in college. Such information may influence the decision to fund a consumer's college education in preparation for a career, rather than placement and training directly into the workforce. Additionally, students who have faced academic challenges within the K- 12 school system, may have significant reservations about the possibility of being successful in college. Negative selfperceptions and academic self-efficacy may develop and become an encumbrance to students, for whom attending college is a viable option (Lackaye, Margalit, Ziv, & Ziman, 2006). The current study provides an illustration for SWD, that success in college may be an achievable goal. The results of our study could be used as a psychoeducational tool to positively impact academic self-efficacy.
The results of the intercept and slope as outcome model suggest that SWD are likely to achieve higher grades in online courses. These findings run contrary to the implication of the intercept as outcome model, which indicated that the average student - not accounting for disability status - is more than likely to achieve higher scores in traditional courses. Despite the implication that SWD fare better in online courses, this is not indicative of poor performance in traditional settings. Based on our findings, we do not sponsor a mass exodus of SWD from traditional courses to online. Yet, this finding may encourage continued learning among individuals whose physical and environmental challenges (e.g., limited public transportation in a rural area) have historically hampered post-secondary options (Crudden, Sansing, & Butler, 2005; Purdie & Boulton-Lewis, 2003). Recognition of online learning as a viable option to the traditional courses (based on achievement) may also help to eliminate some reservations students may have about attending college via the internet.
Students alleviate disability related academic challenges by recognizing individual learning style and developing strategies to compensate (e.g., seeking help when necessary, self-regulation of learning and time management, matching learning style with course type and teaching style). In the vein of empirical data, it may be feasible that SWD performed better in online courses because of the multimodal nature of relaying online course content, which benefits a broader spectrum of learning styles (Englert, Yong, Dunsmore, Collings, & Wolbers, 2007; Hecker, Burns, Elkind, Elkind, & Katz, 2002; Trainin & Swanson, 2005). By contrast, SWOD often prefer learning through written examples and explanations (Heiman, 2006), which may explain why the average student in the current study scored higher in traditional courses, which are mostly taught in such a fashion. Yet, utilizing technology in traditional classes has been established as more efficacious than traditional lectures alone (Buckley & Smith, 2007; Michael & Trezek, 2006; Yesilyurt & Kara, 2007). With this in mind, university personnel may be encouraged to consider ways to integrate more multimedia technology (e.g., Smart Classrooms, Blackboard, WebCT, PowerPoint, Tegrity) into the classroom. Doing so may have broad implications for the greater learning community, not just SWD.
While the results of this study may generate a sense of optimism regarding SWD, consumers of this research should be mindful of the limitations. Learning styles are not universal among those with disabilities, and it is quite feasible that level of online achievement among SWD (within disability) might vary based on disability type. Despite high levels of academic achievement among students with visual and perceptual disabilities (i.e., Blind/visually impaired, Deaf/hard of hearing), these students may encounter significant technical problems accessing course materials (Epp, 2006; Hart, 2008; Saumure, 2004). Similarly, students with LD and ADD/ADHD comprise the largest group of college students with disabilities. Yet, it is not clear what has limited this group's ability to graduate at rates that are commensurate to its growth in attendance. Future research should look to the within disability effects on grades to gain a better understanding of this relationship. In addition to recognizing within disability effects, future researchers may consider accounting for the use of assistive technology (AT) during analyses. Assistive technology is any device that assists with working or learning and the technology ranges from low tech (e.g., mechanical pencils, magnifying glass) to high tech (e.g., screen reading computer programs, voice synthesizers). Accounting for AT use may provide more accurate research findings since proper use and availability of accessible course content can have a positive impact on academic achievement (Englert, Zhao, Collings, & Romig, 2005; Hecker et al., 2002).
All analyses in this paper are based on the course grade as outcome variable which requires a normality assumption (i.e., rij is normally distributed) and that the outcome measure should be metrical (i.e., interval or ratio data). Unfortunately, the course grade used in this analysis is possibly ordinal rather than metrical. Investigating online versus traditional classes and student disability with another dependent variable (e.g., dichotomous variable such as dropping out) or with an alternate methodological treatment on the dependent variable (e.g., treat them as ordinal) within context of Hierarchical Generalized Linear Model (HLGM; Raudenbush & Bryk, 2002) would be an alternative methodological choice, and will be the future direction of this research.
The implementation of this study and corresponding results can serve positive uses for both individuals with disabilities and those who serve them (e.g., universities, disability support services personnel, rehabilitation counselors). This substantive work serves as an initial inquiry, examining the impact that multilevel variables have on the effect of disability status on grades. The significant findings for SWD have provided a preliminary confirmation of the existing literature about accessibility to postsecondary education better facilitated through improved technology. We have provided an empirical "starting point" for anyone interested in the academic success of those with disabilities taking online courses. Nonetheless, it is important that others with shared concerns consider the current methodology and findings and continue what we have begun.
Aitkin, M., & Longford, N. (1986). Statistical modeling issues in school effectiveness studies. Journal of Royal Statistical Society: A, 149, 1-43.
Allen, I. E., & Seaman, J. (2006). Making the grade: Online education in the United States 2006. Needham, MA: Sloan-C Publishing.
Americans with Disabilities Act. 1990. Pub. L. 101-336. U.S. Code. Vol. 42 § 12101 et seq.
Aragon, S. R., Johnson, S. D., & Shaik, N. (2002). The influence of learning style preferences on student success in online versus face-to-face environments. The American Journal of Distance Education, 16(4), 227-244.
Aragon, S. R., Shaik, N., & Palma-Rivas, N. & Johnson, S. (1999). Comparative analysis of online vs. face-to-face instruction. In Proceedings of WebNet World Conference on the WWW and Internet 1999 (pp. 581-586). Chesapeake, VA: AACE.
Association on Higher Education and Disability (1997). Guidelines for Documentation of a Learning Disability in Adolescents and Adults [Brochure]. Columbus, OH. Retrieved July 1, 2002, from http://www.ahead.org/ldguide.ht
Bertsch, T. R, Callas, P. W, Rubin, A., Caputo, M. P, & Ricci, M. (2006). Effectiveness of lectures attended via interactive video conferencing versus in-person in preparing thirdyear internal medicine clerkship students for clinical practice examinations (CPX). Teaching and Learning in Medicine, 19(1), 4-8.
Buckley, W, & Smith, A. (2007). Application of multimedia technologies to enhance distance learning. RE.view, 39(2), 57-65.
Burgstahler, S. (2002). Distance learning: Universal design, universal access. Educational Technology Review, 10(1), 32-61.
Carr, S. (2000). As distance education comes of age, the challenge is keeping the students. The Chronicle of Higher Education, 46(23), A39-A41.
Chyung, S. Y (2007). Age and gender differences in online behavior, self-efficacy, and academic performance. The Quarterly Review of Distance Education, 8(3), 213-222.
Crudden, A., Sansing, W., & Butler, S. (2005). Overcoming barriers to employment: Strategies of rehabilitation providers. Journal of Visual Impairment & Blindness, 99(6), 325-335.
Diaz, D. P. (2000). Comparison of student characteristics, and evaluation of student success in an online health education course. Unpublished doctoral dissertation, Nova Southeastern University.
Edmonds, C. (2004). Providing access to students with disabilities in online distance education: Legal and technical concerns for higher education. The American Journal of Distance Education, 18(1), 51-62.
Educational Testing Service (1999). Policy Statement for Documentation of Attention-Deficit/Hyperactivity Disorder (ADHD) in Adolescents and Adults (Revised). Princeton, NJ. Retrieved July 20, 2008, from http://www.ets.org
Educational Testing Service (2001). Guidelines for Documentation of Psychiatric Disabilities in Adolescents and Adults. Princeton, NJ. Retrieved July 20, 2008, from http://www.ets.org
Educational Testing Service (2003). Guidelines for Documentation of Physical Disabilities and Chronic Health Conditions in Adolescents and Adults. Princeton, NJ. Retrieved July 20, 2008, from http://www.ets.org
Educational Testing Service (2007) Policy Statement for Documentation of a Learning Disability In Adolescents and Adults, 2nd Ed. Princeton, NJ. Retrieved July 20, 2008, from http://www.ets.org
Englert, C. S., Yong, Z., Dunsmore, K., Collings, N. Y, & Wolbers, K. (2007). Scaffolding the writing of students with disabilities through procedural facilitation: Using an Internet-based technology to improve performance. Learning Disability Quarterly, 30(1), 9-29.
Englert, C. S., Zhao, Y, Collings, N., & Romig, N. (2005). Learning to read words: The effects of internet-based software on the improvement of reading performance. Remedial and Special Education, 26(6), 357-371.
Epp, M. A. (2006). Closing the 95 percent gap: Library resource sharing for people with print disabilities. Library Trends, 54(3), 411-429.
Frieden, L. (2003). People with disabilities and postsecondary education. Retrieved October 20, 2004, from http://www.ncd.gov/newsroom/publications/ 2003/education.htm
Georgia Tech Research on Accessible Distance Education Project (GRADE), Center for Assistive Technology and Environmental Access (n.d.). Access eLearning Tutorial. Retrieved July 1, 2008, from http://www.accesselearning.net/
Githens, R. P. (2007). Older adults and e-learning: Opportunities and barriers. The Quarterly Review of Distance Education, 8(4), 329-338.
Halberstam, B., & Redstone, R (2005). The predictive value of admissions materials on objective and subjective measures of graduate school performance in speech-language pathology. Journal of Higher Education Policy and Management, 27(2), 261-272.
Hanana, D. E. (1998). Higher education in an era of digital competition: Emerging organizational models. Journal of Asynchronous Learning, 2(1), 66-95.
Harbour, W. S. (2004). Final report: The 2004 AHEAD survey of higher education disability services providers. Waltham, MA: Association on Higher Education and Disability.
Hart, K. (2008, June 19). Access denied: The blind or deaf can feel left behind as the tools of technology advance. The Washington Post. Retrieved June 19, 2008, from http://www.washingtonpost.com
HEATH Resource Center (n.d.). Postsecondary students with disabilities: Recent data from the 2000 national postsecondary student aid survey. Washington, D.C.: The George Washington University, Heath Resource Center. Retrieved February 1, 2005, from http://www.heath.gwu.edu/ files/active/0/2000_student_aid_survey.pdf
Hecker, L., Burns, L., Elkind, J., Elkind, K., & Katz, L. (2002). Benefits of assistive reading software for students with attention disorders. Annals of Dyslexia, 52, 243-272.
Heiman, T. (2006). Assessing learning styles among students with and without learning disabilities at a distance-learning university. Learning Disability Quarterly, 29(Y), 55-63.
Horn, L., & Nevill, S. (2006). Profile of Undergraduates in U.S. Postsecondary Education Institutions: 2003-04: With a Special Analysis of Community College Students (NCES 2006-184). U.S. Department of Education. Washington, DC: National Center for Education Statistics. Retrieved July 1, 2008, from http://nces.ed.gov/pubs2006/ 2006184.pdf
Kim, M. M., & Conrad, C. R (2006) The impact of historically black colleges and universities on the academic success of African American students. Research in Higher Education, 47(4), 399-327.
Kinash, S., Crichton, S., & Kim-Rupnow, W. S. (2004). A review of 2000-2003 Literature at the intersection of online learning and disability. The American Journal of Distance Education, 18(1), 5-19.
Klinkosz, W, Sekowski, A., & Brambring, M. (2006). Academic achievement and personality in university students who are visually impaired. Journal of Visual Impairment & Blindness, 100(11), 666-675.
Knapp, L. G, Kelly-Reid, J. E., Whitmore, R. W, & Miller, E. (2007). Enrollment in Postsecondary Institutions, Fall 2005; Graduation Rates, 1999 and 2002 Cohorts; and Financial Statistics, Fiscal Year 2005 (NCES 2007-154). U.S. Department of Education. Washington, DC: National Center for Education Statistics. Retrieved July 1, 2008, from http://nces.ed.gOv//pubs2007/2007 154.pdf
Lackaye, T, Margalit, M., Ziv, O., & Ziman, T (2006). Comparisons of self-efficacy, mood, effort, and hope between students with learning disabilities and their nonLD-matched peers. Learning Disabilities Research & Practice, 21(2), 111-121.
Lee, Y, Driscoll, M. P., & Nelson, D. W (2004). The past, present, and future of research in distance education: Results of a content analysis. The American Journal of Distance Education, 18(4), 225-241.
Li, C, & Irby, B. (2008). An overview of online education: Attractiveness, benefits, challenges, concerns and recommendations. College Student Journal, 42(2), 449-459.
Maurer, T., Weiss, E. M., & Barbeite, R G (2003). A model of involvement in work-related learning and development activity: The effects of individual, situational, motivational, and age variables. Journal of Applied Psychology, 88(4), 707-724.
Marklein, M. B. (2005, October 19). College gender gap widens : 57% are women. USA Today. Retrieved June 19, 2008, from http://www.usatoday.com
Mbilinyi, L. (2006, August). Degrees of opportunity: Adults ' views on the value and feasibility of returning to school. Retrieved July 1 , 2008, from http://www.degreesofopportunity.org
Michael, M. G, & Trezek, B. J. (2006). Universal design and multiple literacies: Creating access and ownership for students with disabilities. Theory Into Practice, 45(4), 311-318.
Morris, L. V, Wu, S., & Finnegan, C. L. (2005). Predicting retention in online general education courses. The American Journal of Distance Education, 79(1), 23-36.
Murray, C, Goldstein, D. E., Nourse, S., & Edgar, E. (2000). The postsecondary school attendance and completion rates of high school graduates with learning disabilities. Learning Disabilities Research & Practice, 15(3), 119-127.
Murray, C, & Wren, C. T. (2003). Cognitive, academic, and attitudinal predictors of the grade point averages of college students with learning disabilities. Journal of Learning Disabilities, 36(5), 407-415.
Planty, M., Hussar, W, Snyder, T, Provasnik, S., Kena, G, Dinkes, R., et al. (2008). The Condition of Education 2008 (NCES 2008-031). National Center for Education Statistics, Institute of Education Sciences, U.S. Department of Education. Washington, DC.
Purdie, N., & Boulton-Lewis, G. (2003). The learning needs of older adults. Educational Gerontology, 29(2), 129-149.
Raudenbush, S. W., Bryk, A. S., Cheong, Y F, & Congdon, R. T (2004). HLM: Hierarchical Linear and Nonlinear Modeling. Chicago: Scientific Software International, Inc.
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data analysis methods (2nd ed.). Thousand Oaks, CA: Sage Publications.
Retsas, A., & Wilson, J. (1996). The critical thinking abilities of RNs entering university courses. Australian Journal of Advanced Nursing. 13(3), 25-3 1 .
Rubin, E., & Lennon, L. (2004). Challenges in social communication in Asperger syndrome and high-functioning autism. Topics in Language Disorders, 24(4), 271-285.
Saumure, K., Given, L. M. (2004). Digitally enhanced? an examination of the information behaviours of visually impaired post-secondary students. The Canadian Journal of Information and Library Science, 28(2), 25-42.
Seo, Y, Abbott, R. D., & Hawkins, J. D. (2008). Outcome status of students with learning disabilities at ages 21 and 24. Journal of Learning Disabilities. 41(4), 300-314.
Snyder, T D., Dillow, S. A., & Hoffman, C. M. (2008). Digest of Education Statistics 2007 (NCES 2008-022). National Center for Education Statistics, Institute of Education Sciences, U.S. Department of Education. Washington, DC.
Tincani, M. (2003). Improving outcomes for college students with disabilities. College Teaching, 54(4), 128-132.
Trainin, G, & Swanson, H. L. (2005). Cognition, metacognition, and achievement of college students with learning disabilities. Learning Disability Quarterly, 28, 261-272.
Ulmer, L. W, Watson, L. W., & Derby, D. (2007). Perceptions of higher education faculty members on the value of distance education. The Quarterly Review of Distance Education, 8(1), 59-70.
U.S. Department of Education, National Center for Education Statistics. (2007). The Condition of Education 2007 (NCES 2007-064). Washington, DC: U.S. Government Printing Office.
Vocational Rehabilitation Act. 1973. Pub. L. 93-112, U.S. Code. Vol. 29, § 701 et seq. Vocational Rehabilitation Amendments. 1998. Section 508, Pub. L. 105-220, U.S. Code. Vol. 29, § 794d.
Wagner, M., Newman, L., Cameto, R., & Levine, P. (2005). Changes over time in the early postschool outcomes of youth with disabilities: A report of findings from the national longitudinal transition study (NLTS) and the national longitudinal transitional study-2. Menlo Park, CA; SRI International.
Walker, S., Turner, K. A., Haile-Michael, M., Vincent, A., & Miles, M. D. (1995). Disability and Diversity: New Leadership for a New Era. Washington, DC: President's Committee on Employment of People With Disabilities & Howard University Research and Training Center.
Weathers, R. R., Walter, G, Schley, S., Hennessee, J., Hemmeter, J., & Burkhauser, R. V. (2007). How postsecondary education improves adult outcomes for supplemental security income children with severe hearing impairments. Social Security Bulletin, 67(2), 101-131.
Yesilyurt, S., & Kara, Y. (2007). The effects of tutorial and edutainment software programs on students' achievements, misconceptions and attitudes towards biology on cell division issue. Journal of Baltic Science Education, 6(2), 5-15.
James F. Stewart II
Coppin State University
The George Washington University
The George Washington University
James F. Stewart II, M.A., C.R.C., Assistant Professor, Department of Applied Psychology and Rehabilitation Counseling, Coppin State University, 2500 W. North Avenue, HHSB #333, Baltimore, MD 21216.