Robin van Rijthoven

47 Impact of semantics on word decoding 2 and rapid naming, we also found a significant effect of the latter two abilities on both pseudoword decoding efficiency and word decoding efficiency. In other words, we found a relation between semantic knowledge and word and pseudoword decoding efficiency via phonological awareness and rapid naming. The found indirect effect of semantic abilities via rapid naming is in line with Wolf and colleagues (2016). However, we found no significant total or direct effect of semantic knowledge on word decoding and pseudoword decoding. This was not expected based on the triangular reading model claiming that orthographic representations are linked to both phonological and semantic representations (Seidenberg & McClelland, 1989). Especially for children with dyslexia, who experience weak orthographic‐phonological connections (Wimmer & Schurz, 2010), the use of semantic knowledge could be considered commendable. The finding that there was no impact of broad and deep semantic knowledge on word decoding for children with dyslexia is not in line with previous research (i.e., Van Bergen et al., 2014). The effect of semantic knowledge on word decoding is also not commensurate with the lexical quality hypothesis claiming that proficient reading requires redundant lexicalized word representations (Perfetti & Hart, 2002). It can tentatively be explained from the fact that the children with dyslexia are only in the beginning of their reading development with a primary focus on phonological recoding, which is heavily influenced by phonological awareness and rapid naming, explaining the indirect effect we found of these abilities on children’s decoding efficiency. Of course, the present study can only be seen as a first step in uncovering the role of semantic abilities in dyslexic children’s processes of learning to read. It should be acknowledged that the present data were cross‐sectional and that causal conclusions regarding relations between semantic abilities and decoding efficiency measures cannot be drawn. It would be informative to know whether typically developing children to a certain extent show an effect of semantic knowledge as well. Future studies following a longitudinal design comparing children with dyslexia and typically developing readers are needed to arrive at final answers regarding the role of lexical semantic abilities in learning to read. A second point to mention is the large variability of age in the sample. Therefore, age is included as a covariate in the analyses regarding the dependent variable. One could argue that age should be included as a covariate in the analyses regarding the independent variables as well. This would lead to underpowered analyses. Therefore, age is not included on all variables; hence, results should be interpreted with caution. More research with more power is necessary to disentangle the effect of age on the model. Furthermore, it should be considered to use measures of receptive and not expressive semantic abilities instead of measures that require word retrieval and verbal formulation because dyslexia and specific language impairment (SLI) are comorbid developmental