Screening for learning disabilities: development of a multiple deficit model of identification

The focus of this thesis is on the operationalisation and identification of Learning Disabilities (LD) in a classroom setting. Although cognitive processing deficits often form part of the definition of LD, they are rarely utilised in methods of identification (Kavale, Holdnack, & Mostert, 2005). For instance, the discrepancy method identifies students with a significant discrepancy between aptitude and achievement (e.g., Berk, 1983; Ferrer, Shaywitz, Holahan, Marchione & Shaywitz, 2010). The primary aim of this thesis was to pilot a freely available screening tool that can identify possible cases of LD in a classroom setting. The validation of the discrepancy method of identification of LD is often sought by testing the cognitive processing deficits of the students identified. In this study the reverse of this approach is taken so that the convergent validity of a method of identification based on cognitive processing deficits will be assessed with the discrepancy method of identification. Phonological processing is the most widely accepted cognitive processing deficit associated with LD (van der Leij & Morfidi, 2006) and a nonword reading list was developed in this study to test for this. In order to test orthographic processing, which has a more opaque role in LD, an irregular word reading test was developed. Both word lists were refined and psychometric support for their validity was provided through both item response theory and classical test theory. Rasch analysis was used to demonstrate the good model fit of both of the refined nonword and irregular word lists. In order to account for the heterogeneity of LD, naming speed and verbal memory deficits were also used as indicators of the disorder. Two Rapid Automatic Naming (RAN) tests, using letters and colours, and preliminary norms for digit span tests, both forwards and backwards, were also developed. One of the most important attributes of these tests is that they are simple to administer and score, making them well suited to classroom teacher administration. In order to allocate students into discrepancy groups, the Raven’s Progressive Matrices (Raven, Court & Raven, 1998) and Progressive Achievement Test in Reading (Australian Council of Educational Research, 2008), as measures of aptitude and achievement respectively, were also administered to the entire sample. Finally, the South Australian Spelling Test (Westwood, 2005) was administered so that the relationships between spelling and the cognitive processes could be assessed. Participants were sourced from five primary schools in the Eastern and South- Eastern suburbs of Melbourne, Australia. The sample was made up of 172 Grade 3 participants, as well as 58 Grade 1 and 57 Grade 2 students, who provided a reading age-based comparison for poor Grade 3 readers. Two cluster analyses, one with reading and cognitive variables and one with spelling and cognitive variables, indicated that these cognitive processes were more closely related to reading than spelling and as such the focus for the rest of the analyses was focused on readingbased, rather than spelling-based LD. All Grade 3 students were allocated into LD, Low Achievement (LA) or Regular Achievement (RA) groups as dictated by both the traditional and regression-adjusted discrepancy methods. The cognitive processing attributes of these two groups were examined. LA students, on the whole, had the same cognitive processing proficiency as their LD counterparts. The only exception was that, unexpectedly, LD students scored significantly lower on irregular word reading than their LA counterparts when the groups were dictated by the regression-adjusted discrepancy method. Discriminant Function Analysis was used to investigate whether cognitive processes can be used to predict membership in both sets of discrepancy defined groups. In the first DFA, it was found that 77% of participants were correctly allocated into traditional discrepancy defined groups using only scores on nonword reading, RAN and digit span. In the second DFA, 82% of participants were correctly allocated into regression-adjusted discrepancy defined groups using the same three predictor variables. Interestingly, in the regression-adjusted discrepancy DFA no students were allocated into the LA group, despite the presence of poor readers in the RA group. Although there were still no differences between the LD and LA groups as allocated by the DFA model predicting traditional based discrepancy groups in phonological processing or naming speed, the LA group had significantly lower verbal memory scores than the LD group. This indicates that, contrary to the proposition that LD students have cognitive processing deficits that do not trouble an LA cohort; some LA students are facing the same deficits as LD students, in addition to verbal memory deficits that may reduce their aptitude. Furthermore, as verbal memory was utilised as a successful predictor variable in this model, these findings suggest that LD is best represented by a multiple deficits model in which at least three cognitive processing deficits may contribute to the disorder. Finally, a new method of screening for LD students based on cognitive deficits alone is put forward. In this multiple deficits model, students were designated as possibly LD on the basis of reading age matched phonological deficits or chronological aged matched naming speed or verbal memory deficits. Using this method, 79% of students were placed in the same groups as they would have been through the regression-adjusted discrepancy method. As predicted, multiple deficits resulted in more severe reading deficits, despite reading scores playing no role in the allocation into these new groups. By utilising classroom friendly, freely available tests that automatically calculate results and generate student profiles, more students can gain access to screening tools that can predict the possibility of LD and hence enable more appropriate remediation.