effet du nombre des graphèmes en Anglais - Aix Marseille Université

Appendice III 257Graphemes are perceptual reading unitsArnaud Rey Johannes C. ZieglerCRNC-CNRS, Marseille, France CREPCO-CNRS, Aix-en-Provence, FranceArthur M. JacobsPhilipps-University, Marburg, GermanyGraphemes are currently defined as the written representation of phonemes. For example,the word "BEACH" is composed of three phonemes (i.e., /b/, /i/, and /J/) and consequently,three graphemes (i.e., "B ", "EA ", and "CH "). Graphemes can thus be consideredas the minimal functional bridges between orthography and phonology. In the presentstudy, we investigated the hypothesis that graphemes are processed by the reading systemas perceptual units. We assumed that if the reading system processes graphemes as units,then detecting a letter in a word should be harder when this letter is embedded in a multilettergrapheme than when it corresponds to a single-letter grapheme. In Experiment 1,done in English, participants were longer to detect a letter in a target word, when the letterwas embedded in multi-letter grapheme (i.e., "A" in the grapheme "EA " like in "BEACH")than when it corresponded to a single-letter grapheme (i.e., "A" in "GRASS"). In experiment2, this effect was replicated in French. Together, these results support the view accordingto which graphemes are minimal functional reading units in alphabetic writtensystems like English or French.INTRODUCTIONIn alphabetic writing systems like English or French, letters are considered as the basic distinctive elementsof the word's orthographic representation and phonemes as the basic distinctive elements of the word'sphonological representation (Jakobson, Fant, & Halle, 1952). Each word is thus caracterized along the orthographicand phonological dimensions by a particular sequence of letters and phonemes. Accordingly, readingsuppose to retrieve the word’s sequence of phonemes from the word’s sequence of letters.An examination of the relation between the number of letters and the number of phonemes in a corpus ofEnglish or French words shows that, in a majority of cases, words are composed of more letters than phonemes.As an example, the word BEACH is composed of five letters but only three phonemes (/BiJ/). Oneway to surmount the letter/phoneme mismatch is to introduce the notion of grapheme (Coltheart, 1978 ;Venezky, 1970). A grapheme is defined as the written representation of a phoneme (Berndt, LynneD’Autrechy, and Reggia, 1994 ; Berndt, Reggia, and Mitchum, 1987 ; for a discussion, see Henderson, 1985).Following this definition, there is a one-to-one relation between graphemes and phonemes (i.e., the numberof graphemes in a word equals the number of phonemes) and consequently, graphemes can correspond either toa single-letter or a group of letters. In our example, the word BEACH is composed of one single-letter grapheme,B -> /b/, and two multi-letter graphemes, EA -> /i/ and CH -> /J/. Therefore, graphemes, instead ofletters, appear as the minimal functional bridges between orthographic and phonological codes.Graphemes are incorporated at different levels in recent computational models of visual word recognition.First, the dual-route model, proposed by Coltheart, Curtis, Atkins, and Haller (1993), suggests that a letterstring is processed by two parallel routes : a lexical and a non-lexical route. In the lexical route, the lettersequence activates in parallel lexical entries through an whole-word matching procedure. In the non-lexicalroute, the letter sequence is serially segmented into graphemes in order to perform a sub-lexical orthographyto-phonologycomputation. Second, the parallel and distributed processing model, proposed by Plaut, McClelland,Seidenberg, and Patterson (1996), suggests that graphemes are used as orthographic units at the orthographiclevel, in order to perform the orthography-to-phonology mapping. In both models, graphemes appearas crucial elements of the reading system.This research was supported by a grant from the French Ministry of Education and Research (#95124, programmeSciences Cognitives) to A. Reyand by a grant from the German Research Foundation (DFG : Deutsche Forschungsgemeinschaft)to Arthur Jacobs. Correspondence should be addressed to A. Rey, Centre de Recherche enNeurosciences Cognitives, CNRS, 31, Chemin Joseph Aiguier, 13402 Marseille Cedex 20, France (e-mail :frog@lnf.cnrs-mrs.fr).