The central components of the model correspond to three potentially independent linguistic spelling routes. Two of these, the lexical-semantic route and the lexical-nonsemantic route, are used for spelling familiar words. In contrast, plausible spellings for unfamiliar words or pronounceable nonwords (e.g., sprunt) are assembled by the nonlexical route. An additional central component that receives and temporarily stores the abstract orthographic representations computed by the three spelling routes is referred to as the graphemic buffer.
Spelling by the lexical-semantic route relies on interactions between the semantic system and the graphe-mic output lexicon (Fig. 2). The semantic system represents conceptual knowledge of word meanings independent of word forms. The graphemic output lexicon contains information about the orthographic structure of familiar words and thus functions as the memory store of learned spellings. As depicted in Fig.
2, semantic input to orthography may be direct (pathway A) or indirect via the phonological representation of the word that is also used in speech production (pathways B and C). It has been proposed that representations in the graphemic output lexicon are normally activated by combined input from both the direct and the indirect routes. Such dual coding may safeguard against errors that might occur if one or the other route was used exclusively. For instance, relying on the indirect route via the phonological output lexicon may result in homophone confusions (e.g., "stair"-"stare") since these words have the exact same sound pattern even though they are spelled differently. On the other hand, spelling by the direct route might be susceptible to semantic errors (e.g.,
Figure 2 Cognitive information processing model of spelling and writing.
Figure 2 Cognitive information processing model of spelling and writing.
"day"-"week") unless the response was constrained by simultaneous phonological input to orthography via the indirect route. Postulating dual access to the graphemic output lexicon may also provide an explanation for the clinical observation that some aphasic patients with damage to the phonological output lexicon can write words they are no longer able to say (by relying on the direct route to the graphemic output lexicon via pathway A).
As can be seen in Fig. 2, the lexical-semantic route provides the only mechanism for incorporating meaning into writing. Therefore, this spelling route plays a critical role in conceptually mediated writing tasks such as written composition and written naming. It is likely that writing familiar words to dictation is normally performed via the lexical-semantic route as well, although the lexical-nonsemantic route can also potentially be used for this purpose.
Spelling to dictation by the lexical-nonsemantic route relies on connections between the spoken and written forms of the same word that bypass the semantic system. One possibility is that the dictated word first activates its representation in the auditory input lexicon, followed by the retrieval of the corresponding entry from the phonological output lexicon, which in turn activates the appropriate orthographic word form in the graphemic output lexicon (Fig. 2, pathways D and C). An alternative mechanism might involve direct connections between corresponding representations in the auditory input and the graphemic output lexicons. As it will become apparent later, the evidence for lexical-nonsemantic spelling comes mostly from patients with agraphia, and it is not entirely clear what function this spelling route might serve under normal circumstances.
Normal individuals can produce plausible spellings for unfamiliar words or nonwords without significant difficulty. Since these novel items are not represented in the graphemic output lexicon, spelling cannot simply rely on the activation of stored orthographic patterns. According to our model, the spelling of unfamiliar words and nonwords is accomplished via the nonlexical spelling route. Unlike the lexical spelling routes that rely on a whole-word retrieval process, spelling by the nonlexical route involves a subword-level algorithmic procedure based on phoneme-grapheme conversion rules (Fig. 2). In this process, the novel auditory stimulus is first broken down into its component sounds, following which each constituent phoneme is converted into the corresponding grapheme. It is possible, however, that the nonlexical route can also perform phonological-to-orthographic translations based on units larger than individual phonemes and graphemes (e.g., syllables). Finally, it has been suggested that spelling nonwords may not be entirely nonlexical and may instead be based on lexical analogy with familiar words.
Phoneme-grapheme conversion plays an important role in learning to spell, but this rule-based nonlexical procedure is resorted to much less frequently once the normal adult spelling vocabulary is established. However, the nonlexical route can serve as a backup strategy when word-specific orthographic information is temporarily unavailable or is incomplete. This might happen when normal individuals attempt to spell words they do not use very often (i.e., low-frequency words). It should be noted that in orthographically opaque languages such as English, in which sound-to-spelling relationships are notoriously inconsistent, the nonlexical procedure can only succeed with unambiguous or regular words that have highly predictable phoneme-grapheme correspondences (e.g., "mint"). Ambiguous words, in which the same phonology can be realized by more than one combination of letters
(e.g., "drain" and "drane"), and irregular words that contain exceptional phoneme-grapheme mappings (e.g., "choir") cannot be spelled correctly by the nonlexical route since the straightforward application of phoneme-grapheme conversion rules for such words will result in phonologically plausible errors (e.g., "yot" for "yacht"). The spelling of ambiguous and irregular words, therefore, depends critically on access to precise word-specific orthographic knowledge.
Central spelling routes compute abstract orthographic representations that can be externalized in writing, oral spelling, typing, or as an arrangement of anagram letters. The graphemic buffer is a working memory system that temporarily stores these abstract orthographic representations while they are being converted into codes appropriate for the various output modalities (e.g., letter shapes for written spelling or letter names for oral spelling). As shown in Fig. 2, this processing module receives the output of all three spelling routes and therefore occupies a strategic position between the central and peripheral components of the writing process. The maintenance of information within the graphemic buffer is influenced by stimulus length (i.e., the number of graphemes that make up the spelling of a word or nonword) since longer items require more storage capacity than shorter ones. Dysfunction of the buffer is associated with the loss of information relevant to the identity and serial ordering of stored graphemes, leading to letter substitutions, additions, deletions, and transpositions.
Was this article helpful?