91 Role of semantics in a phonics through spelling intervention 4 Predicting individual differences in phonological and orthographic spelling Research question 1b was about the role of semantic abilities in predicting the phonological andorthographic spellingof childrenwithdyslexia. Toanswer thisquestion, hierarchical regression analyses were conducted on all three outcome measures (i.e., phonological errors, morphological errors, and orthographic errors). Given that we had a control group that was matched on age, we chose to use raw scores instead of standardized scores as it would allow us to compare impact of each measure. In step 1, phonological awareness, rapid automatized naming, and verbal working memory were entered, followed by semantic abilities in step 2. Semantic abilities was added separately to see the unique contribution of semantic abilities beyond phonological awareness, rapid automatized naming, and verbal working memory. Correlations between all used measures are presented in Table 4.2. The results of the analyses (see Table 4.3) showed that rapid naming and verbal working memory predicted morphological errors (Step 2). In Step 2 the effect of rapid naming on morphological errors was no longer present. Higher scores on semantic abilities resulted in fewer phonological, morphological, and orthographic errors and higher scores on verbal working memory resulted in fewer morphological errors. Table 4.3 Cognitive Predictors of Spelling Errors in Children with dyslexia (n = 52). Phonological errors Morphological errors Orthographic errors ∆R2 b ∆R2 b ∆R2 b Step 1 .151 .268** .168* Phonological awareness -2.711 -2.965 -3.045 Rapid automatized naming 2.738 4.751** 3.031 Verbal working memory -7.083 -7.986* -7.467 Step 2 .123** .275*** .143** Phonological awareness -1.071 -0.478 -1.208 Rapid automatized naming 1.194 2.411 1.302 Verbal working memory -6.312 -6.816* -6.602 Semantic abilities -2.803** -4.250*** -3.140** Total R2 adj .211** .503*** .250** *p< .05. **p< .01 ***p< .001 To check whether age had any impact on the results, we reran the hierarchical regression analysis including age in step 1 and the predictor measures in steps 2 and 3; results remained the same.