35 Impact of semantics on word decoding 2 stronger pressure can be hypothesized to make finer phonological distinctions in the mental lexicon. It can thus be assumed that lexical representations start out holistic and become more specified during early andmiddle childhood (Metsala &Walley, 1998). Indeed, it has been found that children’s degree of lexical specificity enhances their phonological awareness (Garlock et al., 2001; Van Goch et al., 2014) and facilitates their decoding skills (Elbro et al., 1998). Following a dual route perspective on reading, it can be hypothesized that a more specified lexicon fosters the process of word decoding and visual word recognition. According to the dual route model (see Coltheart et al., 2001), children learn to assign phonology to new words (or pseudowords) by applying the alphabetical principle. In doing so, they store orthographic, phonological, and semantic knowledge in their mental lexicon. Gradually, they become able to address this stored information in the lexical retrieval of frequently encountered words and in a direct route of visual word recognition (Coltheart, 2006). It can thus be expected that semantic knowledge may foster word decoding indirectly via lexical retrieval (i.e., rapid naming) or directly via word recognition. Neurocognitive research has indeed evidenced that poor readers rely to a greater extent on their semantic lexicon when it comes to word decoding (Shaywitz et al., 2003). Children with dyslexia may thus compensate for their weak orthographic and phonological representations by using their broad and deeply developed semantic knowledge when combining graphemes and phonemes to a recognizable meaningful word. Because the lexical representations of semantic knowledge and phonological knowledge seem to be reciprocally connected, a broad and deep semantic knowledge can be considered useful when phonological representations are less developed (Li et al., 2004). Behavioural evidence for the role of semantic abilities in word decoding has also been found by Nation and Snowling (2004), who showed that weaker semantic skills related to lower word decoding skills, and Ouellette and Beers (2010) who found semantic skills to predict decoding in Grade 6 and irregular word recognition in Grades 1 and 6. There is also evidence that children with dyslexia compensate for their poor decoding skills by using the semantic context during text reading (Nation & Snowling, 1998). Furthermore, a study of Van Bergen and colleagues (2014) showed verbal IQ (defined by expressive vocabulary, expressive syntax, and comprehension) to be uniquely related to later monosyllabic word decoding in 4‐year‐old children who go on to develop dyslexia. Besides a direct effect of semantic representations on pseudoword and word decoding, also indirect effects have been shown in the literature. Swanson and colleagues (2003) showed that semantic skills predicted phonological awareness and rapid naming as well as word decoding. In a similar vein, Torppa and colleagues (2010) showed that receptive and productive semantic skills in Scandinavian children at familial risk for dyslexia predicted their word decoding through phonological awareness, letter naming, and inflectional morphology.