Robin van Rijthoven

43 Impact of semantics on word decoding 2 To answer the research questions, two regression based mediation analyses were conducted with the process add‐on in SPSS (Hayes, 2013). Bootstrapping was set at 5.000 cycles, as recommended by Hayes (2013). In mediation models, the addition of the direct effect (c′) and the indirect effect (ab) form the total effect (c) of a certain independent variable on a dependent variable. The indirect effect (ab) is formed by multiplying the effect of the independent variable on the mediator (a) and the effect of the mediator on the dependent variable (b). A total effect may not reach significance, evenwhen the direct effect is significant (Hayes, 2009), which could be due to a relatively small sample size (Shrout & Bolger, 2002). The final model of this paper includes two statistical models with different dependent variables. In the first model, word decoding was the dependent variable, and in the second model, pseudoword decoding was the dependent variable. In both analyses, verbal working memory, perceptual organization, and age were added as covariates. To ensure that the analyses had enough statistical power, the covariates were added in the models of the dependent variable only. Therefore, the direct and indirect effect do not add up to the total effect. All regression weights can be found in Table 2.3 The results of the first mediation analysis with word decoding as the dependent variable showed no significant direct effect of semantic abilities (c’ = 0.0255, SE = 0.0248, 95% CI [−0.0316, 0.0825]; p = .374) nor a total effect of semantic abilities, (c = 0.0414, SE = 0.0319, 95% CI [−0.0226, 0.1055]; p = .200). Semantic abilities, however, was significantly related to Phonological awareness (a = 0.1712, SE = 0.0615, 95% CI [0.0478, 0.2945]) and Rapid automatized naming (a = −0.1879, SE = 0.0657, 95% CI [−0.3197, -0.0560]) and Phonological awareness (b = 0.1099, SE = 0.0427, 95% CI [0.0240, 0.1985]) and Rapid automatized naming (b = -0.1238, SE = 0.0409, 95% CI [-0.2061, -0.0416]) were significantly related to word decoding. Indirect effects of semantic abilities were found via Phonological awareness (ab = 0.0188, SE = 0.0115, 95% CI [0.0037, 0.0524]) and Rapid automatized naming (ab = −0.0233, SE = 0.0109, 95% CI [0.0066, 0.0495]). The total indirect effect of semantic abilities was significant (0.0421, SE = 0.0173, 95% CI [0.0162, 0.0856]). The relation between semantic abilities and word decoding was partially mediated by Phonological awareness and Rapid automatized naming. The total effect of the model is R2 = 0.691 (MSE = 0.3052, p < .001). Age had a significant total effect on word reading (0.0485, SE = 0.0086, 95% CI [0.0313, 0.0657]). There were no significant effects of covariates perceptual organization and verbal working memory.