An examination of reading, reading development and disorder in a highly transparent orthography: the case of Turkish

Abstract

The primary focus of the current research program concerns visual word recognition and reading aloud as a function of orthographic transparency to inform current debates about the nature of visual word recognition. Within this thesis, this topic is explored using several different approaches with Turkish as the medium of choice. Additionally, the extreme orthographic transparency of Turkish serves as an excellent medium to test theories of visual word recognition. Any universal framework would need to account for the variation found in the writing systems of the Turkic family. Using a computational linguistic method, Chapter 2 explores current definitions of orthographic transparency and novel means of quantifying orthography, extending this approach to Turkish. The result was the production of four language models that take into account the different phoneme inventories used in Turkish, as well as the two main strategies (Word-onset vs whole-word), employed to investigate the quantification of Turkish. The models produced stipulate that Turkish is more transparent than any other alphabetic orthography that has been quantified to date. The chapter also highlights the superiority of whole-word approaches in capturing a full range of variation within an orthography despite some of the current cross-linguistic limitations of using such a method. Chapter 3 examines the currently available resources for Turkish psycholinguistic research and in response to the discovery of a lack of resources in the domain, has led to the creation of the Turkish Lexicon database. The new resource is a sizeable psycholinguistic database that includes several measures of word frequency, contextual diversity and orthographic neighbourhood density as well as providing lexical information such as word and syllable length, bigram and trigram frequency. The Turkish Lexicon was validated using a lexical decision task and also produced a subcorpus for use in future psycholinguistic studies with children. Furthermore, there has been hardly any empirical research investigating linguistic, metalinguistic, and cognitive processes involved in reading the highly transparent orthography of Turkish. To address this, Chapters 4 and 5 investigate how these skills might impact Turkish children who are learning to read and also aims to uncover how developmental dyslexia might manifest itself in Turkish. As such, the current research has the potential to add to our understanding of the cognitive mechanisms that underlie reading in alphabetic languages. It is envisaged that the findings of this study will add to the current debate concerning the distinct influence of universal principles and specific variations in writing systems on cognitive reading processes. In addition, the research will provide conceivably the most comprehensive account of typical and atypical reading development in Turkish-speaking children to date which has huge potential practical implications. Chapter 4 examines the reading strategies of monolingual Turkish schoolchildren while they completed both a single-word naming and oral reading fluency task amongst a battery of cognitive tasks. The findings of the rapid development of phonology as well as the use of two distinct strategies in single-word reading lend support to the weak versions of the phonological and orthographic depth hypothesis of reading. Chapter 5 continues to pursue this question by examining reading disorder, i.e., Developmental Dyslexia in Turkish children. According to Wydell and Butterworth's (1999) Hypothesis of Transparency and Granularity, transparent orthographies such as Turkish should not manifest with a high incidence of phonological dyslexia. The findings of Chapter 5 lend support to this position as well as being in line with multiple deficit models of dyslexia (Pennington, 2006). In Chapter 6, the behavioural data of this doctoral thesis is supplemented with the development of a computational model of visual word recognition in Turkish, the first of its kind. The model builds on the recent incorporation of a self-teaching algorithm (Pritchard, 2012) in the Dual Route Cascaded model of reading aloud and word recognition (Coltheart, Rastle, Perry, Langdon, & Ziegler, 2001). Simulations include exposing the model to vocabularies of varying size to simulate different stages of vocabulary growth in reading development. In addition, Chapter 6 took preliminary steps in investigating the manifestation of developmental dyslexia in Turkish from a computational perspective.