(d) "Emic LRs". Under this scheme, differences between input and output are factored out for a common channel - i.e. for the auditory channel, graph"emes". Thus there will be just one LR for "cat"- perhaps the phonemic form /kæt/ - serving both hearing and speech. This allows the lexicon to capture commonalities in this channel. Similarly, there will be a graphemic lexicon serving both reading and writing. I have proposed a subtype of this model in which there is a "phonological lexicon" serving hearing and speech, mapped into a "semantic lexicon" serving hearing and speech, and can be extended to include a graphemic lexicon mapped into the semantic lexicon and into the phonological lexicon.
- Butterworth (2004), a pag.26