jal wrote: ↑Tue Aug 15, 2023 4:21 am
Starbeam wrote: ↑Sat Aug 12, 2023 12:34 pmBecause i agree with the idea that standard English /ŋ/ is basically /ng/ ([ŋg]), with the /g/ dropping at the end of a syllable, unless a vowel follows in the next.
I invoke Occam's razor here. No need for such a convoluted explenation. Diachronically yes, /ŋ/ came from /ŋg/, but synchronically there's no reason to maintain it's not a phoneme (the biggest evidence being "singer" etc. nog having the /g/).
I agree with this here - it is generally better to posit simpler phonological rules than more complex ones when they adequately predict the surface forms.
Of course, I have found cases in the English I am familiar where neither simple phonological rules nor more complex ones seem to work. Simple rules based on simple tests like being able to form minimal pairs (e.g. if minimal pairs were a rock-solid test, my dialect of English should have phonemic vowel length) quickly run into issues (e.g. vowel length would still be allophonic in final syllables without obstruents in their codas, which questions whether there would be phonemic vowel length in the first place).
But the same goes for complex rules which attempt to predict a wide range of realized forms from a narrow range of idealized underlying forms (e.g. why is the /t/ in
tomorrow or
tonight liable to be flapped, or why is the /b/
able to (or one or both /b/'s in
probably), the /v/ in
over or
every, or the /ð/ in
other liable to be elided, or like, when there is no real way to create phonological rules that predict these particular cases from idealized underlying forms, and furthermore trying to encode these in the underlying forms themselves causes problems with the coherence of the phonological system that is posited, e.g. eliding the /v/ in
every causes major problems because the resulting /ɛr/ does
not merge with the /ɛr/ in
merry because the /ɛ/ in
every is centralized whereas the /ɛ/ in
merry is not).
I personally have not found any good solution to this dilemma myself, even though I have thought about it way too much. The best solution I have thought of is that words have both underlying forms and realized forms
and these need not match each other; the more frequently the words are used the more likely these are to differ, whereas the underlying forms still govern things such as rhyming and phonotactic rules w.r.t. loanwords and coinages.
