Creating versatile morphemes?

[Osverai]

So, I'm trying to get my conlang's grammar worked out, and at the moment, I'm fiddling with case and gender. One snag I keep hitting is the worry that if I create a case ending/marker/affix/etc., it will sound good with one word, or set of words, but not with another. For instance, if I wanted my nominative case ending to be "-asa", that would sound fine applied to a word like "carn", but not so well with a word like "cas". I'm not quite well-versed enough with the structure of case and the like to find some logical way around this. Any help?

[CrazyEttin]

So, I'm trying to get my conlang's grammar worked out, and at the moment, I'm fiddling with case and gender. One snag I keep hitting is the worry that if I create a case ending/marker/affix/etc., it will sound good with one word, or set of words, but not with another. For instance, if I wanted my nominative case ending to be "-asa", that would sound fine applied to a word like "carn", but not so well with a word like "cas". I'm not quite well-versed enough with the structure of case and the like to find some logical way around this. Any help?

You could just have irregular declension for words which you think sound bad with the standard affixes.

[Mardigny]

So, I'm trying to get my conlang's grammar worked out, and at the moment, I'm fiddling with case and gender. One snag I keep hitting is the worry that if I create a case ending/marker/affix/etc., it will sound good with one word, or set of words, but not with another. For instance, if I wanted my nominative case ending to be "-asa", that would sound fine applied to a word like "carn", but not so well with a word like "cas". I'm not quite well-versed enough with the structure of case and the like to find some logical way around this. Any help?

I second CrazyEttin; I would take a historical/phonological approach, if that would appeal to you more. Essentially, you could imagine a 'proto-form' like *cas-asa and then maybe rhoticize one 's' or the other, creating *carasa or *casara, and explain the irregularity as historical dissimilation.

[CrazyEttin]

So, I'm trying to get my conlang's grammar worked out, and at the moment, I'm fiddling with case and gender. One snag I keep hitting is the worry that if I create a case ending/marker/affix/etc., it will sound good with one word, or set of words, but not with another. For instance, if I wanted my nominative case ending to be "-asa", that would sound fine applied to a word like "carn", but not so well with a word like "cas". I'm not quite well-versed enough with the structure of case and the like to find some logical way around this. Any help?

I second CrazyEttin; I would take a historical/phonological approach, if that would appeal to you more. Essentially, you could imagine a 'proto-form' like *cas-asa and then maybe rhoticize one 's' or the other, creating *carasa or *casara, and explain the irregularity as historical dissimilation.

You could actually create different declension categories, eg. all words with stems ending in -sa get a different nominative affix. This way you wouldn't have to create much irregularity if you don't want to, only few additional rules.

[Systemzwang]

Haplology could be quite well motivated there:
cas-asa -> casa

Alternatively, you could have some kind of stem partially distinct from the unmarked form (need not be distinct for every noun, and there is a wide space of options available), so you could get

cas -> cad- -> cad+asa = cadasa

(c.f. Finnish vekotin - vekottime|n, although this is from a historical *vekotim afaict where #m->#n, and this change didn't analogize into inflected forms. OTOH, there's also other things in Finnish like, a syllable that is turned from open to closed needs to reduce its initial consonant a bit along a certain hierarchy, so pata -> padat. alus -> aluksessa rather than alusessa, due to #ks -> #s, I surmise, pikkunen -> pikkus|en for whatever odd reason, etc. )

Another option, depending on how your language works, could be something like caasa, or cassa or whatever. Maybe a: doesn't exist, but e: does, so you get ce:sa whenever a long "a" would be expected, whatever. Languages do a lot of strange things along those lines.