Sound changes occur unconditionally?

[vehke]

Various sources mention that sound changes occur unconditionally, that is without regards to the phono-morphological boundaries or grammatical functions, even though I continuously stumble upon counterexamples.

For example, at some point both Estonian and North S?mi lost word-final nasals for nominal inflection, yet the verbal suffix -n (1sg) remained. Both the nominal nasals and the verbal -n occurred in the proto language.
The verbal -n might have remained under pressure of losing an important semantic distinction, but both languages could resort to pronouns without being forced to innovate.
I have read speakers of some Dutch dialects (and even the literary language) pronounce the final nasal of plural nouns, but not the final nasal of plural verbs. Actually, I remember a grammar handbook of Dutch commenting prescribing this phenomena for the pronunciation of the literary language.
(My dialect retains word-final -n.)
How about Romance languages? It seems both the nominal and verbal morphology was affected by sound change from the little I know, and most languages lost the morphological case distinction.

Is there perhaps a tendency for morphemes in (heavily) agglutinative or polysynthetic languages to be unaffected by sound change or to be immune to loss (of the morpheme as a whole)?

[gach]

For example, at some point both Estonian and North S?mi lost word-final nasals for nominal inflection, yet the verbal suffix -n (1sg) remained. Both the nominal nasals and the verbal -n occurred in the proto language.
The verbal -n might have remained under pressure of losing an important semantic distinction, but both languages could resort to pronouns without being forced to innovate.

I'd suggest it's in deed a matter of loosing semantic distinctions. In both cases the case ending *-n (genitive in E and genitive-accusative in S) caused consonantal gradiation in stem consonants. Look for example the North-S?mi word bahta "back". Its acc-gen case is bađa which differes from the nominative only by gradiation ht:đ. We can reconstruct its original acc-gen as *bađa-n. You see that adding the -n causes the stem consonants ht to lenite to đ. The final -n and thus the whole case ending could easily dissapear still leaving nom and acc-gen cases look different.

This could have happened also with verbs because there also exists consonantal gradiation and also other stem variation and many personal forms in deed look similar. Still I'd say the sg1 -n remained because the three singular persons are the most common and it's important to keep them apart. Because the sg3 hasn't any ending (though having some stem variations) it propably was simpelest to just keep the sg1 -n. N-S?mi still has quite large tendency to leave personal pronouns away every time it's possible and I don't think that tendency has been weaker at the time of this change. That could be why it never complitely relied od pronouns.

Is there perhaps a tendency for morphemes in (heavily) agglutinative or polysynthetic languages to be unaffected by sound change or to be immune to loss (of the morpheme as a whole)?

Might be. Changes that that are regular in roots can have very much different patterns in affixes or the cahnges might even be complitely different. For example in colloquial Finnish at Helsinki region there's a change in the 3rd infinitive illative -maan/m??n > -V' thus producing ole-maan > ole-e' "(go somewhere) to be". There the final -n has disapeared leaving -' which meanes that the initial consonant of the following word is geminated (as itself it has no sound and isn't never marked in writing). The remaining part has complitely disappeared leaving only a longer version of the preciding vowel. These changes are complitely unique and exist nowhere else. Even better is that with verbs with stem ending in a long vowel sound the only change is that the final -n usually disappeares and doesn't even leave ane final gemination, sy?-m??n > sy?-m??. This particular verb form is very common which explains these complex change patterns.

[gsandi]

Various sources mention that sound changes occur unconditionally, that is without regards to the phono-morphological boundaries or grammatical functions, even though I continuously stumble upon counterexamples.

For example, at some point both Estonian and North S?mi lost word-final nasals for nominal inflection, yet the verbal suffix -n (1sg) remained. Both the nominal nasals and the verbal -n occurred in the proto language.
The verbal -n might have remained under pressure of losing an important semantic distinction, but both languages could resort to pronouns without being forced to innovate.

The regularity of sound change was a dogma heavily pushed by the 19th century neogrammarians, and I can see why - there was some very bad historical linguistics done before that time, and the notion of regularity and predictablity helped linguistics become more of a science.

But all we can really say is that regular phonetic change is regular - i.e. rely on tautology. For example, initial h- in Latin always disappeared, there are no counterexamples.

Quite often, however, sound change is not regular, and no reasonable conditioning factor can be found to account for exceptions - in some cases, the change is so haphazard that it's hard to say what is the regular change and what is the exception.

A good example is the fate of the Old French diphthong spelled <oi> which ended up in Modern French either as /wa/ (spelled still <oi>) or as /E/ (spelled today as <ai>, although this was also spelled <oi> up to the end of the 18th century). There is simply no good way to state the environment where one change happened and where the other, indeed some words, like the Old French word "fran?ois" could go both ways: /-wa/ as a personal name and /-E/ as the name of the language and people. But you can't even associate the /-E/ development with the semantic range "people, language", because some language names (su?dois, danois) have the /wa/ development.

[Ran]

Various sources mention that sound changes occur unconditionally, that is without regards to the phono-morphological boundaries or grammatical functions, even though I continuously stumble upon counterexamples.

For example, at some point both Estonian and North S?mi lost word-final nasals for nominal inflection, yet the verbal suffix -n (1sg) remained. Both the nominal nasals and the verbal -n occurred in the proto language.
The verbal -n might have remained under pressure of losing an important semantic distinction, but both languages could resort to pronouns without being forced to innovate.
I have read speakers of some Dutch dialects (and even the literary language) pronounce the final nasal of plural nouns, but not the final nasal of plural verbs. Actually, I remember a grammar handbook of Dutch commenting prescribing this phenomena for the pronunciation of the literary language.
(My dialect retains word-final -n.)
How about Romance languages? It seems both the nominal and verbal morphology was affected by sound change from the little I know, and most languages lost the morphological case distinction.

Is there perhaps a tendency for morphemes in (heavily) agglutinative or polysynthetic languages to be unaffected by sound change or to be immune to loss (of the morpheme as a whole)?

I think the general rule is that sound changes can't occur with grammatical function as a factor, that is, we can't have all t's change into d's in adjectives only. But sound changes do seem to occur to one morpheme at a time, e.g. to, say, the nominative -ak to change into -a, and that might make the accusative -ek to change into -e, but not all -k's.

[vohpenonomae]

I think the general rule is that sound changes can't occur with grammatical function as a factor, that is, we can't have all t's change into d's in adjectives only. But sound changes do seem to occur to one morpheme at a time, e.g. to, say, the nominative -ak to change into -a, and that might make the accusative -ek to change into -e, but not all -k's.

Even this principle is sometimes violated for unknown reasons. E.g., PA /m/ changed to /n/ in some Cheyenne formal nouns, as in mahkwa 'bear' > nahkohe. But it's hard to find an example of this outside of formal nouns; and it's an irregular change on top of that.

Admittedly, such things don't occur often, and conlangers would be good not to rely on them too much.

[zompist]

You should all follow the regularity hypothesis, even though it's not true.

Sound changes over the long run are nearly regular, which is why the hypothesis is good practice. If the Neogrammarians hadn't taken it seriously, no one would have noticed that the exceptions to Grimm's Law follow Verner's Law, and Saussure wouldn't have come up with his coefficients sonantiques, the most charming example of prediction in linguistics.

Data from dialect linguistics and Labov's investigation of ongoing sound change show, however, that an incipient sound change is quite sporadic. The change spreads to more and more words and may eventually affect every word with its phonological conditions-- so from a perspective centuries down the road it may look regular. But some exceptions may remain.

(A possible ongoing sound change: in Valley Girl talk, as represented by Moon Zappa, 'know' is pronounced [nEu]. So far this may affect only a few words (I think she pronounces 'oh' as [Eu] too). If it continues we could expect more and more /o/ to be affected, seemingly at random, until all or mostly all instances have changed.)

Personally, I think analogy accounts for much more sound change, morphological change, and even synchronic morphology, than it's given credit for. Since sound change is sporadic, a model where an entire phoneme shifts over is not realistic. Analogy does the job, however: other words start to shift, one by one, to match the models. Analogy is also an alternative way of storing rules. E.g a French speaker might have a rule to assign gender to new foreign borrowings... or she might just assign them by looking at a similar-seeming word.

[vehke]

Personally, I think analogy accounts for much more sound change, morphological change, and even synchronic morphology, than it's given credit for. Since sound change is sporadic, a model where an entire phoneme shifts over is not realistic. Analogy does the job, however: other words start to shift, one by one, to match the models. Analogy is also an alternative way of storing rules. E.g a French speaker might have a rule to assign gender to new foreign borrowings... or she might just assign them by looking at a similar-seeming word.

As far I understand, we should rather follow the regularity hypothesis because with analogy these diachronical changes would have to be detailed on a case-by-case basis, right?

If these changes happen sporadic and by analogy, I assume the speaker can limit the perceived phonemic/phonetic deviation -- as uttered by other speakers of the community -- by becoming conscious of the grammatical environments, mainly word or morpheme class, so that a sound change does not occur regularly or not at all with a grammatical class it does not match with. Do you think there are any conditions for determining whether there would be the need for such a limit, i.e. important semantic distinctions, lack of circumlocutions, or do you think a sound change (or any other change for that matter) would depend on the speaker's choice of purely phonemic analogy by disregarding the grammar?

Now I've come to realize -- sound change would freeze if the speaker has no memory of the past generations' sound changes, making analogy across grammatical boundaries impossible ('how to tell whether such-and-such word went through a particular sound change or not'). I'll have to put some more thought into this.

From a conlang/conworld POV, it might be an interesting experiment on a small scale to derive dialects and idiolects through a speaker's analogy with utterances as heard in the community, instead of homogeneous shift. Cumbersome, even tiresome yet.

(Hope this post makes sense though. To me even -- when I read it back in the morning.)

[zompist]

As far I understand, we should rather follow the regularity hypothesis because with analogy these diachronical changes would have to be detailed on a case-by-case basis, right?

Not quite: you should follow the regularity hypothesis because that way you'll make consistent, plausible conlangs. A few exceptions are fine, and add flavor, but the bulk of your vocabulary should follow regular rules.

If these changes happen sporadic and by analogy, I assume the speaker can limit the perceived phonemic/phonetic deviation -- as uttered by other speakers of the community -- by becoming conscious of the grammatical environments, mainly word or morpheme class, so that a sound change does not occur regularly or not at all with a grammatical class it does not match with. Do you think there are any conditions for determining whether there would be the need for such a limit, i.e. important semantic distinctions, lack of circumlocutions, or do you think a sound change (or any other change for that matter) would depend on the speaker's choice of purely phonemic analogy by disregarding the grammar?

I'm not sure I'm following. Most sound change (and for that matter analogy) doesn't seem to be conscious at all. If people do become conscious of an ongoing change, they usually deplore it. (And often end up adopting it when they're not paying attention.)

[vehke]

As far I understand, we should rather follow the regularity hypothesis because with analogy these diachronical changes would have to be detailed on a case-by-case basis, right?

Not quite: you should follow the regularity hypothesis because that way you'll make consistent, plausible conlangs. A few exceptions are fine, and add flavor, but the bulk of your vocabulary should follow regular rules.

I don't see why. Detailing an internal history where a sound change only occurs in, say, 30% of the vocabulary (mostly common words), shouldn't affect plausibility. This is not the same as completely arbitrary sound changes.

If these changes happen sporadic and by analogy, I assume the speaker can limit the perceived phonemic/phonetic deviation -- as uttered by other speakers of the community -- by becoming conscious of the grammatical environments, mainly word or morpheme class, so that a sound change does not occur regularly or not at all with a grammatical class it does not match with. Do you think there are any conditions for determining whether there would be the need for such a limit, i.e. important semantic distinctions, lack of circumlocutions, or do you think a sound change (or any other change for that matter) would depend on the speaker's choice of purely phonemic analogy by disregarding the grammar?

I'm not sure I'm following. Most sound change (and for that matter analogy) doesn't seem to be conscious at all. If people do become conscious of an ongoing change, they usually deplore it. (And often end up adopting it when they're not paying attention.)[/quote]

No, I don't sound change happens consciously either, but surely extending the change throughout one's idiolect does? But my original question was whether a speaker is conscious of the grammatical categories in his/her language when 'applying' sound changes or not, by analogy with words that have been uttered in community and have already been mutated.
If the speaker is, a certain sound change could occur in only a small set of words of a particular category, say pronouns, while all other words remain unaffected.

But will they deplore it even if a change is occurring in their own community? I'm not too sure about the negativity towards language change. Of course, it happens a lot in 'prescriptive' (literary) language, where change isn't always appreciated. So I'm mainly talking about the language as spoken by a community, not the national language.

[bi's sock puppet]

Grammatical categories are a somewhat artificial concept; we don't "know" right from the beginning that there are such things as "nouns" and "verbs", we know these things because we learnt them from linguists. So it's not very natural for one to be "conscious" of word classes. My guess is, if a sound change only affects some classes of words, it may in fact have been conditioned by prosody.

[pharazon]

Grammatical categories are a somewhat artificial concept; we don't "know" right from the beginning that there are such things as "nouns" and "verbs", we know these things because we learnt them from linguists. So it's not very natural for one to be "conscious" of word classes. My guess is, if a sound change only affects some classes of words, it may in fact have been conditioned by prosody.

I wonder how the conditions on French liaison fit into that. IIRC, it doesn't happen across major constituent boundaries (the thread explaining where seems to be gone...), but are there notable prosody distinctions between the situations where liaison occurs and where it doesn't?

[vehke]

Grammatical categories are a somewhat artificial concept; we don't "know" right from the beginning that there are such things as "nouns" and "verbs", we know these things because we learnt them from linguists. So it's not very natural for one to be "conscious" of word classes. My guess is, if a sound change only affects some classes of words, it may in fact have been conditioned by prosody.

I would say to become "conscious" of a particular word class would imply that the speaker applies certain morphological process to a particular word class (such as verbs), and not to an entire different word class (such as nouns). These concepts are not artificial, because they're apparently morphologically and perhaps semantically distinct. In other words, I think a speaker very well knows to which class a word belongs, otherwise one could basically attach any affix to any root regardless whether it's a verb, noun, pronoun, &c (that is not to say such things don't happen).
Hypothetical example: if a particular language has a set of subject pronouns which passivize a sentence, I'd say a speaker would still identify it as belonging to a class of pronouns, even though a linguist would conclude the pronoun also encodes verb-related information -- which would then be an artificial concept, I suppose.
In a language that does not have morphologically distinct adverbs and adjectives, thus where those two artificial linguistic concepts belong to the same word class, well that would be a different matter. In that case, I doubt whether the speaker will be "conscious" of them.

Prosody itself takes grammatical (and morphological and semantical) category into account, or not? Isn't that what happens in the French liaison that pharazon mentions?

[vohpenonomae]

Here's what I suspect:

A sound change (or a certain set of them) is operational in a language over a finite time period; oftentimes this period is long enough for a phoneme or sequence to be completely transformed via analogy from sporadic changes. Somtimes, however, the sound change's "time limit" expires before this happens; that accounts for a lot of irregular changes.

[Siride]

Grammatical categories are a somewhat artificial concept; we don't "know" right from the beginning that there are such things as "nouns" and "verbs", we know these things because we learnt them from linguists. So it's not very natural for one to be "conscious" of word classes. My guess is, if a sound change only affects some classes of words, it may in fact have been conditioned by prosody.

Don't tell that to a generative grammarian or...well, you don't want to know.

[vehke]

Here's what I suspect:

A sound change (or a certain set of them) is operational in a language over a finite time period; oftentimes this period is long enough for a phoneme or sequence to be completely transformed via analogy from sporadic changes. Somtimes, however, the sound change's "time limit" expires before this happens; that accounts for a lot of irregular changes.

So, how come there is such a time limit in the first place?

[vohpenonomae]

Here's what I suspect:

A sound change (or a certain set of them) is operational in a language over a finite time period; oftentimes this period is long enough for a phoneme or sequence to be completely transformed via analogy from sporadic changes. Somtimes, however, the sound change's "time limit" expires before this happens; that accounts for a lot of irregular changes.

So, how come there is such a time limit in the first place?

Tough to say generally. It may be a generational thing in some cases; an older generation of speakers dies off, leaving a sound change incomplete, and the new generation doesn't carry it on or they find their own changes that supercede the old ones.

[Salmoneus]

Or maybe the sound change is promulgated by a popular few. If those few disappear or become unpopular, the sound change may stop, or even reverse.
An extreme example would be a revolution. A sound change starts when people imitate the court speech, and gets partway into the language. But then there's a revolution, and the new peasant rulers reject "bourgeois" speech patterns, and attempt to remove the sound change. But in some words, the change is old enough that they don't remember that it came from the court speech, so it stays.
Of course, that is, as I say, an extreme example.

[zompist]

Not quite: you should follow the regularity hypothesis because that way you'll make consistent, plausible conlangs. A few exceptions are fine, and add flavor, but the bulk of your vocabulary should follow regular rules.

I don't see why. Detailing an internal history where a sound change only occurs in, say, 30% of the vocabulary (mostly common words), shouldn't affect plausibility.

Can you give some examples of sound changes that affect only common words in any natlang?

The general experience of historical linguists is simply that sound changes make no reference to commonness of words, or to syntactic categories. It doesn't mean it's impossible, but it does mean that any such proposal needs to be triple-checked.

But will they deplore it even if a change is occurring in their own community? I'm not too sure about the negativity towards language change. Of course, it happens a lot in 'prescriptive' (literary) language, where change isn't always appreciated. So I'm mainly talking about the language as spoken by a community, not the national language.

It has nothing to do with literary or national languages. If you go to a Quechua village and try speaking a different form of Quechua than the local dialect, they'll strongly correct you.

My point is just that if a sound change is consciously noticed at all, the reaction is likely to be negative.

[Diogenes]

...
Data from dialect linguistics and Labov's investigation of ongoing sound change show, however, that an incipient sound change is quite sporadic. The change spreads to more and more words and may eventually affect every word with its phonological conditions-- so from a perspective centuries down the road it may look regular. But some exceptions may remain.

(A possible ongoing sound change: in Valley Girl talk, as represented by Moon Zappa, 'know' is pronounced [nEu]. So far this may affect only a few words (I think she pronounces 'oh' as [Eu] too). If it continues we could expect more and more /o/ to be affected, seemingly at random, until all or mostly all instances have changed.)
...

Does that mean that analogy is responsible for the regularity of sound change, if not for the sound change itself?

[vehke]

Can you give some examples of sound changes that affect only common words in any natlang?

Really depends on what constitute common words: a lot of Dutch speakers have intervocalic /d/ > /j/ in |goede| /xuj@/ "good" (colloquially also spelled |goeie|) and |beneden| /b@nej@/ "downstairs" (and a few others, |rijden| "to ride", |rode| "red"). There are dialects (apparently) which have completed this sound change, however, most speakers I've heard only show this phenomenon with a fixed set of words.
Perhaps someone with a formal knowledge of Dutch and its dialects can shed some light on this phenomenon.

It has nothing to do with literary or national languages. If you go to a Quechua village and try speaking a different form of Quechua than the local dialect, they'll strongly correct you.

Of course, I wouldn't expect the local speaker to embrace sound change from outside of their community. Surely it would be different it that were the case for a group of speakers from within the community? Well, not surely, but that's what I would imagine especially if those speakers are respected.
Perhaps literary languages have nothing to do with it, however I don't think literary languages bind its speakers as a community.

[Siride]

The sound change of /e/ > /o/ or /u/ following a /w/, a very old one, still hasn't affected all possible words. So I have "woman" [wUm@n] and "twenty" [twUni], but "women" [wImIn] and "went" [wInt].

[Siride]

Can you give some examples of sound changes that affect only common words in any natlang?

The general experience of historical linguists is simply that sound changes make no reference to commonness of words, or to syntactic categories. It doesn't mean it's impossible, but it does mean that any such proposal needs to be triple-checked.

"says" and "does" versus "plays" and "moos".

[Diogenes]

The sound change of /e/ > /o/ or /u/ following a /w/, a very old one, still hasn't affected all possible words. So I have "woman" [wUm@n] and "twenty" [twUni], but "women" [wImIn] and "went" [wInt].

Excuse me, please, if I am confused, but is this an answer to my question? I think, that it's quite fundamental, so it would be fine to be answered, or at least discussed.

[Siride]

The sound change of /e/ > /o/ or /u/ following a /w/, a very old one, still hasn't affected all possible words. So I have "woman" [wUm@n] and "twenty" [twUni], but "women" [wImIn] and "went" [wInt].

Excuse me, please, if I am confused, but is this an answer to my question? I think, that it's quite fundamental, so it would be fine to be answered, or at least discussed.

It was supposed to be in response to Zompist's question about sound changes related to commonness, though in retrospect I didn't indicate that (and my examples are haphazard). So, that's what it is.

Sorry about just randomly inserting some sound changes with no introduction to the purpose of the post.

EDIT: This isn't even your thread, so I'm not particularly sure why the outburst. It is possible to have multiple discussions in a single thread. Just because you make a post and add a new question doesn't mean that you monopolize the topic of the thread.

[Diogenes]

Reading over some entries, I think these things are related.

If a sound change starts with a word or two, and spreads over similarly sounding ones, this is analogy, isn't it? It is in this case a phonological analogy, not a grammatical one, but an analogy nevertheless, if I'm not too much mistaken.

If that is true, we've fixed analogy as the driving force of constituting a language, or a dialect, or whatever , especially if we find it working on syntax, too.

If that is not fundamental for any language, I) don't know what might be fundamental.