The process of learning to read is crucial to understand the meaning of written text. A large part of reading research is devoted to identifying the components that facilitate the acquisition of reading skills (e.g. Patel et al., 2004, Ziegler et al., 2010). These studies are directed at discovering, to a great extent, the predictors of reading in alphabetic scripts. This paper attempts to elaborate in further detail these predictors of reading acquisition in the context of cross linguistic research.
In an alphabetic orthography, the characters correspond to separate phonemes. It is now generally accepted that reading acquisition also depends to an extent on the orthographic depth of the language. Based on the consistency of letter-sound mappings, alphabetic orthographies are classified into two kinds - deep or opaque orthographies and shallow or transparent orthographies. For example, in a transparent orthography like German, the letter sound mappings are highly consistent, where as in English, which is a deep orthography; each letter may represent different sounds depending on the context. The orthographic depth hypothesis posits that word recognition processes vary depending on the orthographic depth of the language (Katz & Frost, 1992). Katz and Frost hypothesized that word recognition in shallow orthographies involves phonological representations whereas in deep orthographies, visual-orthographic coding that taps the internal lexicon is involved. Results from a cross language experiment of English, Hebrew and Serbo-Croatian ,studying reaction times to words and non-words, found that the naming tasks which require phonological recoding were faster than lexical decision tasks in Serbo-Croatian, which is the shallowest orthography of the three (Frost, Katz & Bentin, 1987). Further research comparing literacy acquisition in a number of European orthographies revealed that development of decoding skills is slowest in inconsistent orthographies such as English (Seymour, Aro & Erskine, 2003). The above studies provide evidence in support of the orthographic depth hypothesis which suggests that orthographic depth plays an important role in the development of reading in alphabetic orthographies. Also, from the evidence that the processes of reading are different in different orthographies, it can also be suggested that the predictors of learning to read may also vary depending on the depth of the language.
One of the predictors of learning to read is phonological awareness, which is the ability to identify and manipulate individual sounds in spoken language (Yopp, 1992). A recent cross linguistic study was conducted by Ziegler, Bertrand, Toth, Csepe, Reis and Faisca (2010), across five different languages varying in orthographic depth (Finnish, Hungarian, Dutch, Portuguese and French). The study attempted to investigate the role of phonological awareness, memory, vocabulary, rapid naming, and non verbal intelligence in reading performance in these languages. The researchers used language specific measures that elicited standardised z scores which allowed for comparison of performance levels between languages. Results from a large sample of grade 2 children showed that universal predictors of reading are consistent across the alphabetic scripts, though the degree to which they contribute to reading appears to depend on the depth of the orthography. The study investigated three claims with respect to the role of phonological awareness in learning to read in different alphabetic orthographies. The first claim, that phonological awareness is most important in all languages, and the second claim, that phonological awareness is equally important in all languages, was not supported by the results. Results showed that phonological awareness is a significant predictor of reading across the 5 languages but its influence is stronger in opaque orthographies. A similar view was adopted by Landerl and Wimmer (2000), who drew evidence from studies on German and English children with dyslexia to conclude that phoneme awareness deficits were present only in early stages of reading in the German dyslexic children. This was attributed to the consistent letter sound mappings of the German orthography.
To test these claims, a study conducted on older children learning to read Czech, a transparent orthography, and English, an opaque orthography, found that in normal readers, phonemic awareness was an equally important predictor of reading and spelling ability. Also, dyslexic children in both languages had phonemic awareness deficits (Caravolas, Volin & Hulme, 2005). In a similar vein, a study investigating reading skills in 6-11 year old English and Dutch children found that the predictors of reading, in particular phoneme awareness, were similar in both transparent and opaque orthographies (Patel, Snowling & de Jong, 2004). The authors of both these studies attribute their findings to their use of stringent and reliable parallel speeded measures of phonemic awareness (phoneme deletion and spoonerism tasks) in both the languages. Perhaps longitudinal studies using similar response time tasks of phoneme awareness are necessary to confirm the findings. To address this issue, a longitudinal study was conducted on Greek primary children at the age of 7 and 9 (T1), and after a year at 8 and 10 (T2) (Nikolopoulos, Goulandris, Hulme & Snowling, 2006). Results showed that phoneme awareness was a concurrent predictor of reading, however when T1 reading rate was controlled, only speech rate emerged as a longitudinal predictor of reading. This evidence can be used to support the argument that phoneme awareness may not be the strongest predictor of reading skills in transparent orthographies.
A recent theoretical development that may have different implications for the predictors of reading across orthographies is the psycholinguistic grain size theory (Ziegler & Goswami, 2005). According to this theory, the grain size (size of the linguistic unit) lexical representation of a language depends on its consistency. This in turn results in the development of different reading strategies in different languages. In light of this theory, Georgiou, Parrila andÂ PapadopoulosÂ (2008) proposed that for consistent orthographies, readers rely on grapheme-phoneme recoding strategies which tap phonological short term memory. On the other hand, readers in inconsistent orthographies develop more 'flexible unit size recoding strategies' which requires skills of phonological awareness and orthographic processing. This is similar to the orthographic depth hypothesis stated by Katz and Frost (1992). To test this proposition and to contradict the 'universal predictors of reading skills' view put forth by Patel et al., a longitudinal study measuring the predictors of reading fluency and decoding skills of Greek and English children was carried out (Georgiou et al., 2008). The researchers found the results to be in favour of the grain size theory i.e. it was found that phonological awareness and orthographic processing better predicted reading skills in inconsistent than consistent orthographies. However, this discrepancy, as admitted by the authors, could be because the tasks on the both the languages were not properly matched.
It appears that the use of different phonemic awareness measures in different studies, for cross linguistic comparisons, is the root cause of such varied findings. Since children learning to read consistent orthographies reach ceiling on phoneme awareness tasks faster than children learning to read inconsistent orthographies (Seymour et al., 2003),studies that used phoneme awareness tasks that were relatively difficult and had a speed component found phoneme awareness to be a universal predictor of reading development (e.g. Caravolas et al., 2005). However, this conclusion still appears to contradict a number of single language studies in consistent orthographies (e.g. de Jong & van der Leij, 2003). From the above instances, it can be suggested that phonemic awareness does play a significant role in predicting reading skills in alphabetic languages, but there is no conclusive evidence on the extent of this role.
In recent years, there has also been considerable interest in the role of RAN as a predictor of reading in various consistent and inconsistent orthographies. RAN comprises of timed tasks consisting of stimuli (colours, objects, letters and digits) that are presented to participants for rapid identification. A study conducted by Landerl, Wimmer and Frith (1997) comparing the reading abilities of English and German dyslexic children found that German dyslexic children had relatively higher levels of reading accuracy when compared to controls than the group of English dyslexic children. However, both groups were found to have relatively low levels of reading speed. These results led researchers, de Jong and van der Leij (1999) to propose that deficit in reading speed, and not reading accuracy, is the source of reading problems in consistent orthographies.
In order to identify the role played by naming speed, it is first important to understand the processes involved. Most studies examining the predictive role of RAN in reading highlight the divide in the theories, of the precise part it plays in reading acquisition (e.g. Patel et al, 2004, Georgiou et al, 2010). While one view holds that RAN is the speed by which phonological information stored in memory can be accessed (Wagner et al., 1994), another view posits that RAN is related to the rate of orthographic print recognition (Wolf & Bowers, 1999). As with phoneme awareness, there seem to be contradictory findings about RAN as a predictor of reading across alphabetic orthographies. A longitudinal study on Dutch children revealed that in kindergarten, rapid naming influenced subsequent reading acquisition. This effect persisted even in Grade 1 independent of phonological awareness (de Jong & van der Leij, 1999). Wimmer, Mayringer and Landerl (2000), encountered similar results in a longitudinal study with German children who were poor readers. Results showed that early naming deficits predicted later impairments in reading fluency. On the other hand, a longitudinal co relational study in English, measuring reading abilities in children from kindergarten through fourth grade found that phonological awareness better accounted for later reading skills than naming speed (Wagner, Torgeson, Rashotte, Hecht, & Burgess et al., 1997). Thus, RAN appears to play a greater role in consistent rather than inconsistent orthographies. This is in line with the notion of RAN as a part of phonological processing (Wagner et al., 1994) since one to one grapheme-phoneme coding (Ziegler & Goswami, 2005) used in consistent orthographies requires faster access to the phonological representations.
On the contrary, results from a few cross linguistic studies showed that RAN plays a weak role in both deep and transparent orthographies. The use of colour and object RAN tasks instead of alphanumeric or letter RAN appear to have contributed to the weak results, since stronger correlations have been found between digits or letter RAN and reading (Wolf, Bally & Morris, 1986). For example, in a study of longitudinal predictors of decoding, reading fluency and spelling in three languages of varying orthographic depth (Finnish, Greek and English), RAN was a significant predictor of fluency only in Greek after it was put in the same model with phonological awareness and letter knowledge (Georgiou et al., 2010).Similarly, in the study by Patel et al. (2004), phoneme awareness was a stronger predictor of reading acquisition in both Dutch and English. The authors of both the above mentioned studies concede that the use of digit or letter naming tasks instead of colour or object would have different implications for the role of RAN. There seems to be a lack of evidence from direct comparisons of the role played by RAN in consistent and inconsistent orthographies.
Letter knowledge has been found to be strong predictor of reading acquisition across all alphabetic orthographies, consistent and inconsistent (Bruck, Genesee & Caravalos, 1997, Georgiou et al, 2010). Bowey (2005) suggests that the early familiarity with letter names allows for a basic understanding of the alphabetic principle i.e. that letter names correspond to individual sounds. To explain the distinction made between letter knowledge and letter-sound knowledge, Foulin (2005) suggests that up to school level, letter knowledge is a better predictor of reading. After children start learning to read, letter-sound knowledge assumes greater importance as a predictor of later reading, as by this time most children master the names of letters. In a study on English children at a familial risk of dyslexia, it was found that control children had better knowledge of letter names than at risk children even at the age of 45 months (Gallagher, Frith & Snowling, 2000). Letter knowledge was found to be the strongest predictor of reading performance even in a transparent orthography such as Finnish (Muller & Brady, 2001). Comparative studies between consistent and inconsistent orthographies reported similar findings (e.g. Georgiou et al, 2010). A longitudinal study comparing cognitive and non cognitive predictors of reading skills in Greek and Canadian kindergartners found kindergarten letter knowledge to be the strongest predictor of Grade 2 reading skills in both the languages (Manolitsis, Georgiou, Stephenson & Parrila, 2009). On the other hand, in a study examining the longitudinal predictors of reading and spelling in Norwegian/Swedish and American/Australian children, letter knowledge in preschool predicted Grade 1 reading only in children learning to read English. However, letter knowledge measured at kindergarten significantly predicted reading in both the U.S./Australian and Scandinavian children (Furnes & Samuelsson, 2010). Similar findings were reported by Mann and Wimmer (2002) in a study on German and American children. While American kindergartners possessed sound letter knowledge, the German children did not display perfect letter knowledge till Grade 1 and 2. The researchers in both these studies suggested that the emphasis on early literacy in children learning English and late onset of formal education in some transparent orthographies is the cause for this discrepancy.
Phonological short term memory, to some extent, has been seen to predict reading acquisition in inconsistent orthographies (Bowey, 2005). A longitudinal study on English speaking children tested them on measures of reading and non word repetition at age 4 and age 8 (Gathercole & Baddeley, 1993). When other factors such as non verbal intelligence and vocabulary were controlled, non word repetition at age 4 was found to be a significant predictor of only one measure of reading at age 8. This measure, the Primary Reading test, was found to be the only reading measure that depends highly on phonological recoding strategies. Furthermore, the cross linguistic study on predictors of reading by Ziegler et al. (2010) found that phonological short term memory strongly correlates with reading and decoding accuracy in Dutch and Portuguese, the strongest correlations being with Hungarian . Analysis of the languages at the beginning of the study revealed that Hungarian is more consistent than either Dutch or Portuguese. Thus, based on the evidence that readers of consistent orthographies make use of phonological recoding strategies, it can be hypothesized that working memory is a better predictor of reading skills in consistent orthographies. Single language studies in consistent orthographies with Dutch (de Jong & van der Leij, 1999) and Greek children (Maridaki-Kassotaki, 2002) also found a significant relationship between verbal working memory and reading skills. Studies on Dutch children with reading difficulties found that these children performed worse than then their reading age peers on working memory tasks (de Jong, 1998). However, whether phonological working memory plays a greater role in consistent than in inconsistent orthographies is an area that needs further research. The lack of comparative research with focus on the role of working memory in reading acquisition in deep and shallow orthographies renders it difficult to support the above hypothesis with concrete evidence. Moreover, it is difficult to distinguish the role of phonological short term memory from phonological awareness in both deep and shallow orthographies because, more often than not, phoneme awareness tasks tap on phonological short term memory as well (Alloway et al., 2004). Therefore, there is a need to devise pure measures of phonological awareness and verbal short term memory to understand clearly the role played by each in the development of reading.
Previous research has shown how certain non-cognitive factors can play an important role in shaping the process of learning to read (e.g. Furnes & Samuelsson, 2010). Biological influences on reading acquisition may be conceived to be more or less uniform across countries (Seymour, 2005). However, educational practices, especially early teaching methods appear to have a major influence on the development of reading. For example, Scotland follows a whole word teaching method (Seymour & Elder, 1986), while in countries like Finland and Greece, a synthetic phonics approach (which concentrates on learning the grapheme-phoneme correspondence) is adopted (Georgiou et al., 2010). In England, early teaching methods are an integration of both the phonics and whole word learning (Wagner et al., 1997). The differences in teaching method may lead to differences in reading acquisition across countries (de Jong & van der Leij, 1999). Seymour (2005) noted that these differences may produce large disparities in studies where languages are being compared. Similarly, the age of formal education and country specific literacy practices also appear to affect the process of development of reading. For example, Mann and Wimmer (2002) noted that literacy instruction is delayed till first grade in German speaking countries, whereas in English speaking countries, pre-reading and home literacy activities are emphasized. As a result, English kindergartners performed better than German kindergartners on phoneme awareness tasks. However, by Grade 1, children from both the countries performed equally on the same tasks. This study underlines differences in pace of reading development in different countries as well as the role non-cognitive factors play in shaping skills necessary for learning to read such as phoneme awareness. Though these factors may not directly influence the process of learning to read, they appear to play a major role in shaping the predictors of reading development, thereby indirectly leading to individual differences in reading. Thus, it is important for cross linguistic research to take into consideration all these factors in comparisons of predictors of reading development across languages.
The present paper attempted to evaluate the findings of cross linguistic research with regard to the predictors of reading acquisition in various alphabetic orthographies. Based on the orthographic depth hypothesis, it was suggested that predictors of reading differ in different orthographies. Evidence from cross linguistic research can be used to conclude that predictors of learning to read are similar across alphabetic orthographies. The degree to which they predict reading acquisition, however, depends on the orthographic depth of the language. Conversely, there also exist studies which lay emphasis on the universal nature of these predictors. Methodological constraints for accurate comparisons appear to be a critical reason for the divergent findings in cross linguistic research. Further research should aim at developing and employing tasks that are reliable measures of reading across languages. Moreover, cultural and educational differences in teaching methods must be taken into account to accurately understand the roles of these predictors. In sum, it can be concluded that predictors of learning to read in alphabetic orthographies need to be understood in context of all the above factors and that extensive comparative research with sound methodology is needed to elicit conclusive evidence.