morphophonemes and allomorphs

Morphophonemes and allomorphs

adapted from a 26 Nov 96 post by Mark Line

Timothy Miller wrote:

I am wondering if Mark's discussion on "morphophonemes" doesn't better fall under the term "allomorphs". In English, there are two pronounciations of "the", which are /Di/ and /D@/. They have complimentary distribution (like allophones), and they are predictable in that /Di/ preceeds a word that starts with a vowel, while /D@/ preceeds a word that starts with a consonant.
Would not the variations in form due to vowel harmony also be allomorphs?

Not if you write the morpheme morphophonemically, as I said, because these (particular) morphemes then all have just one morphophonemic shape -- that's the beauty of treating morphophonemes as separate (sound) units. The alternation is controlled completely by phonological context, so it's part of phonology in the wider sense, and certainly not part of morphology any more than lexis is part of morphology. The role played by morphophonemes in morphology is the same as that played by phonemes, namely a merely lexical role -- you represent the (morpho)phoneme sequences of morphemes in the lexicon, and that is pretty much that.

For English "the", you would posit //DI*// in which //I*// is a morphophoneme alternating phonologically between /i/ and /@/. Similarly, you would posit a morphophoneme //S*// in English, used both for 3rd person singular present verb forms and for plurals, as it happens:

//dOgS*// --> /dOgz/ "dogs"
//kAtS*// --> /kAts/ "cats"

//kIlS*// --> /kIlz/ "kills"
//wOkS*// --> /wOks/ "walks"

If, instead of using morphophonemes, you described these phenomena morphologically, treating /s/ and /z/ as two allomorphs of the plural morpheme, and /s/ and /z/ as two allomorphs of the 3SgPres morpheme, you would still not be describing

when you get /s/ and when /z/ for plurals,
when you get /s/ and when /z/ for 3SgPres verbs,
why it is that the rules behind both of these alternations just happen to be identical, and
why it is that these alternations can be described entirely in terms of phonology -- even though you've gone and posited them as allomorphs, as morphological variants.

Positing a morphophonemic representation of both these morphemes, on the other hand, captures all the phenomena in one fell swoop. If for some reason you're interested in seeing how many different phonemic shapes occur for a morpheme, you can always just resolve the morphophonemic alternations you've described and find out. But if you take a primary route through morphology, you are still left with the need to describe a kind of alternation that is phonological in nature and that is completely predictable without recourse to morphology in the first place. If you don't use morphophonemes to do so, you're missing some very important generalizations. If you do use morphophonemes, then you don't (often) have allomorphs after all, because the morphemes have only one morphophonemic shape.

You can see this from the L2-acquisition perspective, too. Isn't it easier to learn -- once -- how a particular morphophoneme works, without having to learn lists of (utterly non-morphological) allomorphs? For English the hard way would be doable, and maybe even for Turkish (although I wouldn't advise it). But I don't think I'd want to try to learn, say, Finnish without getting a handle on all the various recurring morphophonemes in that language -- I'd be learning seemingly illogical allomorphs out the wazoo, otherwise. And learning Turkish is certainly made easier when you learn about morphophonemes like E2, I4 and T and then simply include them on a par with non-alternating phonemes when learning the affixes and vocabulary -- a practice now supported in some textbooks, in fact.

Return to Conlang-related topics|Back to FAQ page