Wednesday, December 27, 2017

Detail #366: A Morphophonological Quirk

For examples here I'll use cases. However, this applies to any prevalent morphological thing - tense, aspect, volitionality, evidentiality, number, etc - it's just a question of finding a way of applying it.

Now, sometimes, sound changes happen at boundaries of words. An example of this is how the former Finnic accusative case suffix -m has merged with the genitive case -n, due to a sound change that turned all final -m into -n, including stem-final -m. Thus, some words whose nominative form ends in -n have inflected forms with an -m- instead: sydän, sydämen. Historically, there's probably also been a form sydämem, which now comes out as sydämen as well.

Beyond cases, a language can have other affixes, e.g. possessive affixes, various clitics, etc. We can now imagine a situation where a different affix blocks a merger of cases by means of having had a different sound change induced (or just plain prevented it), a case distinction can survive in a limited environment, such as, say, before something analogous to Latin -que and similar. Finnish has -kin serving a similar role as -que, and we can imagine then a different version of Finnish having a change -mk- > -mp-. Then, we'd have a situation where 'a heart (nominative) too' would be 'sydämpin', 'a heart's too' would be 'sydämenkin' and 'a heart (acc) too' 'sydämempin'. The negative version of 'too', -kaan ('not even', 'not ... either', 'neither a/the ...'

Now, as I mentioned, this needn't be a case - could be a volition marker or whatever.

Describing such morphological quirks in tabular form requires some special notation, e.g. some kind of diacritic that serves exclusively to mark the existence of an underlying phoneme that may resurface. For the faux-Finnish example, we can consider m̄ for this role. Now, we could get the following pattern for the word sydäm̄:
case underlying formrealization
nom: sydäm̄sydän
acc, clitic kin:sydäm̄em̄kinsydämempin
gen, clitic:sydäm̄enkinsydämenkin
The thing I find relevant or interesting here is really the distribution of mergers vs. distinction. However, a convenient and succinct way of encoding such things in a morphological table is obviously relevant for descriptive purposes. The approach given above - using arbitrarily redefined diacritics - seems to have one great disadvantage: the requirement of learning to mentally apply the sign. In a short description, this is surprisingly taxing. If you've devoted your life to study a particular language, it is no big deal, but in a text you barely read once, it is a bit taxing. Another method that would require a bit more awkward writing, but be more parseable could be something like the following:
lexical example: sydämm > n / _#
morphological example: -emm > n / _#
Essentially replacing m̄ by mm > n / _# throughout the dictionary. Of course, for this use, I assume a really short sample dictionary in a fairly short text.

This could be expanded into marking relevant sound changes wherever they apply in morphophonological contexts.


  1. I'm not sure I understand the representational difficulty. If m > n / _ # is a regular rule, why does it need to be marked in paradigms at all?

    I also find the change *mk > mp to be extremely unlikely. The acoustic cues for place of articulation are seriously attenuated in nasals because of formant damping, meaning that *mk is much more likely to become ŋk than it is to become mp. But I suspect that *mk > mp was just a made up example to illustrate your representational problem.

    I think that an easier (and real) example of what you're after is found in Chemehuevi, a Uto-Aztecan language once spoken on the California/Arizona border. Historically, final vowels were devoiced and then eventually deleted. However, they reemerge upon suffixation. There are minimal pairs: /nɨŋapɨ/ 'chest, ribcage' /nɨŋapi/ 'basket'. Both forms surface as [nɨˈŋap] because of the final deletion rule.

    However, when a possessive pronoun is suffixed to the noun, the original vowel resurfaces: [nɨˈŋapɨn] 'my chest', [nɨˈŋapin] 'my basket'. I think all of the elements of your original example are there: the neutralization of an original distinction (deletion of distinct final vowels) which reemerges in certain contexts. Their representation doesn't have to be particularly complicated. The underlying form for chest is /nɨŋapɨ/, and that for basket is /nɨŋapi/. So if I've understood you correctly, the problem you talk about isn't really a problem. If I've misunderstood, some clarification would be welcome.

    1. This was mainly a bit of a musing with some rather unclear objectives in mind; 1. The point is not phonemes that re-emerge in some positions, which is a fairly common thing in languages (doesn't even, say, French, have that?) - it's about (entire, distinct) morphemes that re-emerge in some positions, not just even morphemes that become distinct in different positions. IIRC I haven't seen this in any conlang ever, but I am sure there exist natural languages with this. Part of the point of this blog is to showcase ideas to conlangers, so I feel justified in including an idea like this.

      The notation was meant to solve one problem I have: when reading about morphophonology, it's pretty hard to memorize all the salient changes, and so having something that reminds one of them when relevant - and in a form that is fairly concise - can be helpful.