effet du nombre des graphèmes en Anglais - Aix Marseille Université

Appendice I 233according to a consonant, vowel, consonant (CVC) or, more precisely, an onset, nucleus, coda (ONC)scheme. This scheme is both motivated by current linguistic theory (e.g., Wiese, 1996), and by empiricaldata (Dell, 1988). We chose this subsyllabic organization as a pragmatic, parsimonious compromise. Higherinfralexical phonological structures, like onset-rime units, for example, are less flexible and less general.Lower structures, like ungrouped phoneme strings, for example, complicate the connectivity between letterand phoneme units.The phonemic representation level thus contains onset, nucleus and coda positions. These units can codesingle phonemes or phoneme clusters. Furthermore, a "silent phoneme" is added at the onset and coda positionsin order to represent monosyllabic words that have no pronounced consonant at their beginning or end.As an example, consider the stimulus GOOD. This is phonologically coded as /gUd/ and decomposed as :Onset = /g/ ; Nucleus = /U/ ; Coda = /d/. Similarly, BLUE (/blu/) is decomposed as : Onset = /bl/ ; Nucleus= /u/ ; Coda = /*/ (the star representing the silent phoneme in the coda position). Table 2 shows thethree sets of phonological units of the MROM-P : They include all possible phonemes or combinations ofphonemes contained in the phonological lexicon of the model at each position.Table 2 : Description of the three sets of phonological unitsONSET51 unitsNUCLEUS34 unitsCODA78 units*, b, bl, br, d, D, dr, dw, f, fl, fr, g, gl,gr, h, J, k, kl, kr, kw, l, m, n, p,pl, pr,r, s, S, sk, skr,skw, sl, sm, sn, sp, spl,spr, Sr, st, str, sw, t, T, Tr, tr, tw, v, w,z, _1, 2, 3, 4, 5, 6, 7, 8, 9, #, $, @, E, i, I,j#, j$, j1, j3, j5, j7, j8, j9, jE, ji, jQ, ju,jV, j{, Q, u, U, V, {*, b, d, D, dst, dT, Dz, dz, f, fT, ft, g, gz,h, J, k, ks, ksT, kst, m, md, mf, mp, mps,mt, mz, kt, l, lb, ld, lf, lk, lm, ln, lp, lS,ls, lt, lT, lv, lz, l_,n, N, nd, ndz, nJ, Nk,Nks, nS, ns, nt, nT, nts, nz, nZ, Nz, n_,p, ps, pt, pT, R, s, S, sk, sCONNECTIONS BETWEEN LETTERS AND PHONOLOGICAL UNITS. Figure 5 illustrates our connectionscheme for linking orthographic and phonological units, a tentative solution to the problem of discoveringan optimal grapheme-to-phoneme correspondence (GPC) scheme for IA-type models. In order to beconsistent with the spreading activation (system) principle of IA models, we carried out an exhaustive analysisof the present lexicon, recording all existing links between each letter/grapheme -at each relative position-and each phonological unit -at the ONC positions. The results were stored in large matrices that codethe GPCs, such as the fact that B in initial position (like in BLUE) is connected to the phonological unit/bl/ at the onset position. Similarly, for the same word, L at the initial +1 position is connected to /bl/ inonset, U at the final-1 is connected to /u/ in nucleus, and so on. Thus, respecting the spreading activationmechanism of the model, when a letter like B is activated in initial position, it will send excitation to allcorresponding phoneme units in onset position, i.e., /b/, /bl/ and /bR/.Figure 5 : Connectivity between the letter and phonological unit levels