The problems of reading, even at the levels of the sentence and extended text, stem largely from difficulties in identifying words. Recognizing a written word entails a mapping from the script to a representation in the reader's internal lexicon. This mapping or decoding is not a natural characteristic of language acquisition (as is the mapping from the spoken work to its lexical representation); rather, it entails specific metalinguistic abilities that rely on careful instruction. Two types of decoding have been suggested. One is direct and related to visual aspects of the printed work, the other is mediated and involves a translation into a phonological representation prior to lexical access. Aside from a few recent exceptions, the majority of the research conducted with English suggests that fluent readers use only the direct visual route. In contrast, research with Serbo-Croatian suggests that the mediated route is routine in fluent reading. The proposal is directed at pinpointing phonology's role in word recognition by examining in detail the similarities and differences between the data obtained with two writing systems. A number of experiments and simulations are planned. They fall into five main groups represented by the following specific questions: (1) Is there an additional and independent process of direct visual access in Serbo-Croatian? (2) Do the early bottom-up processes in English and Serbo- Croatian word recognition differ qualitatively or quantitatively? (3) What kinds of processing units, interacting in what kinds of ways, are needed to model the main factors of word recognition in Serbo-Croatian, and what minimal modifications to these units and interactions would be needed to accommodate the main facts of word recognition in English? (4) What are the dynamics of the processes linking letter units and phoneme units in Serbo-Croatian word identification, given that the letter units represent the characters of two, partially overlapping alphabets? (5) Do word- specific processes follow rule-like processes in the development of fluent word recognition, or vice versa, and is the ordering the same for both English and Serbo-Croatian? Information processing theory and methods (lexical decision, rapid naming, priming, masking) are used to test these contrasting claims, and the modeling or simulation paradigm is that of parallel distributed processing. Subjects will be adult men and women in Yugoslavia and the United States, bilinguals fluent in reading the English and Serbo-Croatian orthographies, and children from both countries.
| Lukatela, G; Turvey, M T (2000) Do spelling variations affect associative and phonological priming by pseudohomophones? Percept Psychophys 62:196-217 |
