Emma wrote:
Quote:
Why have two levels? Because it should make writing morphological rules easier down the road. Say you have an inflectional prefix |e-| and you want to add it to oskal and get eskal rather than *eoskal (or whatever) even though that's not how it works with other vowel-initial words. In that case, you just order your rules so epenthesis applies at the very end, after all the prefixes and suffixes have been added.
It's easier to illustrate with examples.
Let's take two German words, ['tʰaːkʰ] "day" and ['ʃpaːtʰ] "spar [a kind of mineral]". Both of these take the plural ending |ə|, but although the plural of [ʃpaːtʰ] is ['ʃpaːtʰə], the plural of ['tʰaːkʰ] isn't *['tʰaːkʰə]. Instead, it's ['tʰaːgə]. How do you deal with an ending that voices some stops but not others?
Well, if you notice that voiced stops (e.g. [g], [d], etc.)
never occur in final position in German, then you can write a
final devoicing rule. Your underlying forms can have voiced stops in this position but then a phonological rule replaces them with voiceless stops whenever they aren't followed by a vowel.
So then the underlying forms of the word ['tʰaːkʰ] is not */taːk/, since this predicts a plural form *['tʰaːkʰə] which doesn't exist. t's /taːg/. If you apply the final devoicing rule to it, you still get the correct singular form. But if you add the plural ending |ə|, then the rule doesn't apply and you get the correct plural form, ['tʰaːgə].
Your epenthesis rule is like this final devoicing rule. You want your surface forms to have the same sort of phonetic shape, but you want to have more options when you create your word roots. The way it will become clear that this is what you are doing is through alternations like these.
Let me try another example with your made-up words. You had one that was /hena/, right? What happens when you add it to /skal/? You get /henaskal/ because the /s/ has a vowel in front of it so the segmentation of the word is CV.CVC.CVC. All syllables fit the (C)V(C) pattern, so your rule of epenthesis doesn't apply.
This is potentially
different than what happens with /ren/ + /skal/. Say you decide to use /o/ as your default epenthetic vowel. Then putting these together gives you /renoskal/. The segmentation is the same, but now the middle syllable is /nos/ instead of /nas/. Why the difference? Because there is no middle vowel in the underlying form; it's only supplied when needed.