Reading is one of the most important academic skills in our society. However, some children struggle to reach sufficient reading proficiency during and after elementary school education. Between 3 and 8% of the children show serious difficulties in acquiring adequate reading skills (e.g., Landerl & Moll, 2010; Moll, Kunze, Neuhoff, Bruder, & Schulte-Körne, 2014). Difficulties in reading can lead to limited education, resulting in higher chance of unemployment as well as financial and economic disadvantages (Valtin, 2017; Vellutino et al., 1996). Also, there is a higher risk for children with reading difficulties to develop comorbid psychological problems, such as social phobia or depression (e.g., Bäcker & Neuhäuser, 2003; Carroll, Maughan, Goodman, & Meltzer, 2005; Willcutt & Pennington, 2000). To counter such unfavorable development, identification of factors relevant for (un)successful reading proficiency, as well as development and improvement of instruments for recognizing children at risk are crucial (Castles, Rastle, & Nation, 2018; Tippelt & Schmidt-Hertha, 2018).
It is widely accepted that reading includes a variety of processes, which differ in their complexity. These processes include identifying written words, comprehending sentences, and also establishing coherence between sentences (Perfetti, Landi, & Oakhill, 2005; Vellutino, Fletcher, Snowling, & Scanlon, 2004). In this paper, we focus on the processing steps necessary for understanding a text while concentrating on different complexity levels, including basic level (i.e., word-reading) and higher level (i.e., sentence- and text-level comprehension; see Lenhard & Schneider, 2006). The word level represents the basic level of reading, which includes decoding of single words and understanding the meaning of the words (Castles et al., 2018; Karageorgos, Müller, & Richter, 2019). Single word reading is mostly associated with lower level processing skills, such as fast and efficient orthographic and phonological processing and identification (Wolf & Katzir-Cohen, 2001). In an often cited model of reading processes (i.e., Dual Route model; Coltheart, Curtis, Atkins, & Haller, 1993), word reading proceeds in two different ways via (1) lexical or (2) non-lexical route. When confronted with frequent and/or familiar words, representations of words and letter patterns stored in the mental lexicon can be retrieved automatically via the lexical route. When confronted with low frequent and/or unfamiliar words, letter-to-sound-conversion via the non-lexical route is applied. Especially in orthographically consistent languages, such as Greek, Finnish, Italian, or German, children who are learning to read rely more on letter-to-sound recoding because the relationship between letters and sounds is straightforward (Ziegler & Goswami, 2005). Skilled readers are characterized by increased use of the lexical, more efficient route during reading. The automatized word decoding processes free up cognitive resources which can then be used for higher-demanding processes, such as comprehension processes at the sentence and text level (Perfetti & Hart, 2002; Pikulski & Chard, 2005; Roberts, Good, & Corcoran, 2005). Less skilled readers use the non-lexical, less efficient route more often, resulting in decreased reading fluency and therefore impaired reading comprehension (Fuchs, Fuchs, Hosp, & Jenkins, 2001). In addition, at the word level, poor readers show difficulties in their decoding skills compared to good readers (Cain, 2009).
The successful identification of the words, interpretation of their meaning in the given context as well as the integration between different words enable reading comprehension at higher level (i.e., sentence-level; Lenhard & Schneider, 2006). In order to understand the content of the sentence it is necessary to analyze semantic and syntactic elements within the sentence (Richter & Christmann, 2002). The processing of the semantic and syntactic elements is assumed to run parallel, and enables the understanding of the content of the sentence (Lenhard & Schneider, 2006; Taraban & McClelland, 1990). At the sentence-level, poor readers seem to have lower syntactic knowledge and they show more problems in creating a coherent representation of related sentences compared to skilled readers (Cain, 2009).
The last level of reading we want to examine in this study is text-level comprehension, defined as an understanding of the relationship between different sentences, which enables the processing of texts (Gough & Tunmer, 1986; Lenhard & Schneider, 2006). Successful text-level comprehension processes therefore include extraction of information, establishing cross-reference (anaphoric) connections, and also drawing inference conclusions (in a sense of “reading between the lines”; Lenhard & Schneider, 2006). At the text-level, poor readers seem to have problems in building a situational model described in the text due to the problems in generating text-based inferences, and they seem to have problems integrating the propositions from the text with their own prior knowledge (Cain, 2009).
To sum up, efficient word-level reading frees up cognitive resources which are then available for sentence- and text-level comprehension (Perfetti & Hart, 2002). Since word-level reading is mostly associated with lower level processing skills, such as fast and efficient retrieval as well as orthographic and phonological processing and identification, it is important to have a closer look at skills relevant for reading. In this respect, especially the automatized and fast retrieval of information (i.e., naming speed), the quality of orthographic representations and access to linguistic information stored in the mental lexicon (i.e., orthographic knowledge) as well as the awareness of the sound structure of the language (i.e., phonological awareness) have shown to be important for reading operations.
Phonological awareness, naming speed, and orthographic knowledge
Following the assumptions of automatic information processing theories (LaBerge & Samuels, 1974; Perfetti, 1985), phonological awareness, naming speed and orthographic knowledge are considered to be highly interrelated, however, independent contributors to reading. An efficient representation retrieval from the mental lexicon due to established pathways facilitates and accelerates decoding processes, thus, free cognitive resources can be used for higher processes, such as reading comprehension. This fast and efficient retrieval is, however, only achievable, if direct word identification is possible (Thaler, Ebner, Wimmer, & Landerl, 2004) via the lexical route from the mental lexicon. If the direct retrieval is not available, phonological awareness helps to translate print into words via the less automatized and less efficient letter-by-letter identification through phonological re- and decoding processes. However, this impedes the storing of commonly occurring orthographic patterns and representations. As a result, inefficient, slow recognition of orthographic patterns can lead to slower reading rate (Wolf & Bowers, 1999) and difficulties in reading comprehension (Landerl & Wimmer, 2008). As suggested by the orthographic depth hypothesis (Katz & Frost, 1992), the letter-to-sound consistency of the language influences the degree to which readers rely on phonological or orthographic information. In opaque languages (such as English or French) readers rely more on the visual-orthographic structure of printed words to overcome decoding difficulties resulting from ambiguous and partly inconsistent letter-to-sound correspondences. In transparent languages (such as German or Greek) with consistent letter-to-sound correspondences readers rely more on more easily available phonological information (Ise, Arnoldi, & Schulte-Körne, 2014).
Phonological awareness enables the recognition and manipulation of sound units of the spoken language (e.g., Anthony & Francis, 2005; Steinbrink & Lachmann, 2014). Therefore, it is considered as a prerequisite of early decoding development, with an impact on later reading comprehension (Vellutino, Tunmer, Jaccard, & Chen, 2007). A large body of research supports the predictive value of phonological awareness for the acquisition of reading skills (e.g., Ennemoser, Marx, Weber, & Schneider, 2012; Heath & Hogben, 2004; Moll, Fussenegger, Willburger, & Landerl, 2009). A recent meta-analysis by Pfost (2015) identified phonological awareness as a significant predictor for reading speed, reading accuracy, and reading comprehension. However, the differences in letter-to-sound correspondences in transparent vs. opaque languages have been shown to lead to different impacts of phonological awareness on reading (see Ziegler et al., 2010). In transparent orthographies, the differences in preliterate phonological awareness become homogenized more quickly due to consistent letter-to-sound connections than in opaque languages, leading to weaker correlations between phonological awareness and reading in transparent languages (Ziegler et al., 2010). It is assumed that deficits in phonological awareness lead to difficulties in building and storing associations between phonological and orthographic representations in the mental lexicon, thus, evoking difficulties in reading acquisition (Landerl & Thaler, 2006; Scheerer-Neumann, 2015). Children with reading difficulties also show deficits in different aspects of phonological awareness, such as segmentation and phoneme deletion tasks (e.g., “say the word ‘mouth’ without /m/”; Melby-Lervåg, Lyster, & Hulme, 2012; Paulesu et al., 2001; Ramus et al., 2003).
Naming speed is considered as a basic cognitive process which influences reading speed development (Moll, Wallner, & Landerl, 2012). It is usually measured by rapid automatized naming (RAN) tasks requiring the speeded naming of lists of familiar stimuli, such as letters, pictured objects, color patches, and digits (Denckla & Rudel, 1976). Typical tasks used to measure reading at the word-level, such as tasks focusing on word reading speed and/or pseudoword (i.e., pronounceable letter combinations, created by using certain linguistic criteria) reading as well as tasks measuring reading accuracy and speed, are usually correlated with or predicted by RAN measures (e.g., Bowers, 1995; Georgiou, Parrila, & Papadopoulos, 2008; Landerl & Wimmer, 2008; Moll et al., 2009). Furthermore, several studies reported that RAN is also correlated with or predicts reading comprehension measures (e.g., Arnell, Joanisse, Klein, Busseri, & Tannock, 2009; Georgiou, Das, & Hayward, 2008; Kirby, Parrila, & Pfeiffer, 2003; McCallum et al., 2006). It is assumed that in consistent orthographies where readers decode words by applying letter-to-sounds conversion rules (e.g., Goswami, 2002), naming speed is strongly related to reading because the phonological representation of graphemes must be retrieved quickly for letter-to-sound conversion to be effective (Georgiou, Parrila et al., 2008). Children with reading and spelling difficulties show slower naming speed in the RAN tasks compared to unimpaired controls (e.g., Georgiou, Papadopoulos, Fella, & Parrila, 2012; Wimmer, 1993).
Orthographic knowledge is considered as one of the major individual prerequisites of word identification (Cutting & Denckla, 2001), which contains specific sequences of graphemes representing written words (i.e., mental graphemic representations; Apel, 2011). In order to accurately identify a word, it is crucial to know how letters are combined to form a specific word (Apel, 2011; Loveall, Channell, Phillips, & Conners, 2013). Hence, fluent reading is supported by a sufficient level of orthographic knowledge, which enables the individual to quickly recognize written words with little cognitive effort (e.g., Ehri, 2005; 2014). In transparent languages, such us Dutch, repeated blending of graphemes and phonemes during decoding leads to successful phonological recoding, and enables establishment of orthographic representations (De Jong & Messbauer, 2011), which are crucial for developing fluent reading (Rothe, Cornell, Ise, & Schulte-Körne, 2015).
There are several different definitions of orthographic knowledge in the literature. In this paper, we use the term orthographic knowledge consistent with Apel (2011), where both the ability to develop and access mental representations of written words in the mental lexicon (i.e., word-specific orthographic knowledge) as well as the application of knowledge about rules and word patterns (i.e., general orthographic knowledge) are included. Word-specific or lexical orthographic knowledge involves the stored mental representations of familiar words and word parts. Tasks used to measure word-specific orthographic knowledge usually involve words and pseudohomophones decision tasks, where children have to actively think or reflect on their knowledge of orthography and decide whether the presented stimuli is a real word or not by comparing it to the orthographic representation stored in the mental lexicon (e.g., Apel, Henbest, & Masterson, 2018). General or sub-lexical orthographic knowledge contains knowledge about the rules and legal letter patterns of a writing system, including the rules that govern where certain letters can occur in word positions, and what letters can be combined with another (Apel et al., 2018). General orthographic knowledge is typically measured by orthographic word-likeness task, where children are asked to choose the one pseudoword that most reassembles a real word. One pseudoword in these tasks follows legal orthographic conventions whereas the other violates them. For instance, these tasks measure children’s knowledge of permissible orthographic patterns, word position rules, and/or orthographic sequence rules (Apel et al., 2018). With these tasks, the children’s knowledge of allowable orthographic patterns is measured rather than the explicit use of that knowledge during reading or spelling. Children with reading and spelling difficulties show lower scores in the tasks measuring word-specific (Bergmann & Wimmer, 2008) and general orthographic knowledge compared to their typically reading peers (Rothe et al., 2015).
Several studies revealed that word-specific and general orthographic knowledge contribute to word-reading skills (e.g., Arab-Moghaddam & Sénéchal, 2001; Bergmann & Wimmer, 2008; Georgiou, Parrila et al., 2008; Rothe et al., 2015; Rothe, Schulte-Körne, & Ise, 2014). In previous studies examining the role of orthographic knowledge in reading performance, typical tasks used to measure reading skills included real word/pseudoword reading and/or reading fluency, however, without including any higher reading processes, such as reading comprehension (Apel et al., 2018). To our knowledge, there are only few studies examining the relationship between orthographic knowledge and both reading at word-level and sentence- and/or text-level simultaneously in transparent languages with relatively consistent letter-to-sound-correspondences, such as German (e.g., Ise, Arnoldi, & Schulte-Körne, 2014; Zarić, Hasselhorn, & Nagler, 2020). Ise et al. (2014) have shown no significant correlations between orthographic knowledge and reading speed and comprehension in children from kindergarten to 2nd grade. In the reading task used by Ise et al. (2014), children are presented with simple sentences and are asked to read them silently as quickly as possible, and to indicate whether the content of the sentence is generally false or true. In this study, however, only one task measuring orthographic knowledge at the general level (i.e., the knowledge about frequent and infrequent double consonants) was used. Another study conducted in German by Zarić et al. (2020), however, reported that word-specific and general orthographic knowledge show significant correlations with reading at word-, sentence, and text-level. Moreover, the analyses have shown that both word-specific and general orthographic knowledge are significant predictors for reading at word-level, as well as at sentence- and text-level in German 3rd graders without reading difficulties. Since reading accuracy reaches ceiling at the end of Grade 1 in transparent languages (Seymour, Aro, & Erskine, 2003), age differences between the examined samples in these two studies might have an influence on different result patterns. In addition, the differences in the tasks used to measure orthographic knowledge and reading performance might explain differences in the findings.
To sum up, it still remains unclear whether orthographic knowledge with its two components (i.e., word-specific and general) contributes to sentence- and text-level reading in German. Furthermore, the previous studies regarding the contribution of word-specific and general orthographic knowledge at sentence- and text-level reading in German examined typical readers (Ise et al., 2014; Zarić et al., 2020). Therefore, it is still unclear whether both subtypes of orthographic knowledge (i.e., word-specific and general) also contribute to reading at sentence- and text-level in German poor readers.
The present study
Phonological awareness and naming speed have shown to be important predictors for basic as well as for higher reading processes (e.g., Kirby et al., 2003; Landerl & Wimmer, 2008; Pfost, 2015). Word-specific and general orthographic knowledge, on the contrary, were mostly examined in basic reading skills (i.e., word and/or pseudoword reading; e.g., Conrad et al., 2013; Rothe et al., 2015). Regarding higher reading skills (i.e., sentence- and text-level reading), mixed results were reported (Ise et al., 2014; Zarić et al., 2020).
Previous studies in German examining the role of orthographic knowledge in basic and higher reading processes simultaneously included samples of children without reading difficulties. However, studies with different samples (i.e., with and without reading difficulties) report different results (e.g., different correlations for impaired and non-impaired readers) regarding the relationship between reading and components relevant for reading (i.e., phonological awareness, RAN, orthographic knowledge; e.g., Swanson, Trainin, Necoechea, & Hammill, 2003). Therefore, the results from studies examining children without reading difficulties cannot be generalized to children with reading difficulties. Thus, it is important to examine the role of orthographic knowledge to reading in children with reading difficulties as well, as anticipated in the present study. A large body of research also suggests that the course of reading acquisition and the deficits underlying reading difficulties might differ across orthographies (Ziegler & Goswami, 2005). Thus, findings from opaque languages, such as English, cannot be transferred onto more transparent languages, such as German, which is in focus of the present study.
The aim of this study is to address the gaps in past empirical research by exploring the relationship between word-specific and general orthographic knowledge and reading at basic reading level (i.e., word-level), as well as at higher reading level (i.e., sentence- and text-level) in German elementary school poor readers. Since word-specific and general orthographic knowledge seem to play an important role in word-identification, it can be assumed that children with low reading proficiency use their orthographic knowledge when reading single words. Therefore, in line with previous results (e.g., Rothe et al., 2015; Zarić et al., 2020), we hypothesize that word-specific and general orthographic knowledge contribute significantly to reading at word-level, over and above the contribution of phonological awareness and naming speed in poor readers.
Furthermore, following the assumption that high-quality orthographic representations are necessary for higher reading processes, like reading comprehension (see for example the Lexical Quality Hypothesis; Perfetti & Hart, 2002), it is assumed that word-specific and general orthographic knowledge can contribute to efficient single-word reading, enabling their processing, and therefore enhancing reading fluency, resulting in comprehension at sentence- and text-level. Thus, we hypothesize that word-specific and general orthographic knowledge also contribute to reading comprehension at sentence- and text-level, in addition to phonological awareness and naming speed in poor readers.