On 11 Jan 2014 04:57, "Daniel Bowman" <[log in to unmask]> wrote: > The question of how to map our conlang's phonology to individual symbols leads to several solutions, such as: a) Diacretics b) Symbol clusters c) Complete remapping of 'conventional' phonology associated with Roman letters to our new ones as needed or d) Abandonment of Roman script in favor of IPA. If our orthography is intended for ease of reading, options a) and b) are probably easiest. If we are more concerned about explicit phonological precision, options a) is good and d) is best, though I find a) rather intimidating and d) makes the language incomprehensible to anyone not well versed in IPA. I think (d) is worst not only for orthography but also for phoneme symbols. It implies phonemes are defined by realization. It misleadingly ignores dialectal and allophonic variation, and it misleads the reader into thinking the surface pronunciation is evident from the graphical representation. Accordingly I prefer to reserve the use of IPA symbols for phonetics. My roman orthography for Livagian uses only the letters used by Latin, plus overdot and/or underdot, and this serves as a notation for phonological form too. My preferred notation for English phonology uses roman letters, numerals 3 and 6, the euro sign, eth and thorn, and hacheked s,z,c,j. Having said that, there's another conlang of mine that I prefer to write (in my head) using IPA precisely because I want to represent the surface phonetic form rather than underlying phonology and because the surface phonetic form is so complex (using most of the sounds symbolized by the IPA); the only snag there is that Unicode provides no symbol for a retroflex implosive. And.