48 Chapter 2 language disorders (Catts et al., 2005). Besides semantic knowledge also the role of retrieval could be taken into account. A final point to make is that although the results of the factor analysis indicated that the four subtests that measured lexical semantic abilities loaded on one and the same factor, it could be argued that the subtests, similarities, and productive vocabulary, are conceptually more related to semantic abilities as compared with the other two subtests (see Burton et al., 2001; Cohen, 1959). We therefore conducted a secondary analysis with only these two measures (i.e., similarities and productive vocabulary). The overall conclusions remain the same: Semantic abilities were still indirectly related to word decoding and pseudoword reading. The present study leads to implications for future research. It is shown that semantic abilities are indirectly related to word and pseudoword reading, via their relation to phonological awareness and rapid naming. This association could indicate that semantic abilities are a facilitating factor for forming fine‐grained phonological and orthographic representations and so indirectly influence reading development. A recent study by Van Gorp and colleagues (2016) has indeed shown that semantic categorization in addition to feedback and motivation in a word decoding training task may yield important effects on decoding efficiency in poor readers in second grade. To find out to what extent the effects are due to semantic categorization, feedback or motivational aspects of the training are object to future research. Overall, the present findings indicate that the semantic lexicon of children with dyslexia could contribute to pseudoword and word decoding efficiency indirectly. The indirect effects point to the fact that lexical specificity may help dyslexics to become better phonologically aware (see Metsala, 1999) and to become better in lexical retrieval (see Wolf et al., 2016), both of which have a positive impact on decoding efficiency. Furthermore, it fits the lexical restructuring hypothesis in which lexical semantic subsystems could foster the phonological abilities and thereby word and pseudoword reading. Regarding the dual route model, it is possible that the better specified lexicon is the product of the nonlexical route but also partly by semantic development. The finding that there are no direct effects of semantic abilities on decoding measures point to the fact that children with dyslexia do not directly profit from a strong semantic knowledge component in learning to decode words. Based on these results, it seems possible that children with dyslexia compensate their weak ability to form phonological and orthographic representations by use of their semantic abilities as reasoned in the lexical quality hypothesis and lexical restructuring hypothesis. Even though the effect of semantic abilities was small and indirect, these findings show the relevance of a broad and deep semantic knowledge in the reading development of children with dyslexia. Semantic abilities seems to be a variable that should be taken into account in further research regarding reading.