I'm not always the best with sound changes so I've been looking through things like index diachronica to get a handle on the types of sound changes are likely to happen, ranging from one offs with one sound changed in a specific context to a set of phonemes sharing some trait in a set of similar contexts and everything in between. Of course, some more basic sound changes that use general traits of the sounds involved are pretty easy to wrap my head around (voicing unvoiced consonants between vowels, devoicing voiced sounds near unvoiced sounds or at the ends of words, l > ɬ in certain environments), but even going by this there seem to be some odd exceptions in most sets of sound changes.
For example, I saw β ð ɣ → f θ k / _{p,t,k,s}; which means these voiced fricatives become unvoiced before unvoiced plosives+/s/, but /ɣ/ then also becomes a stop. Which isn't too weird, but still maybe not what would be expected. In a grouping like this, is it just more likely for ɣ to do this than fricatives further forward? Is it likely for velars in general to act weirdly in these situations? At least I can set up a general rule that a common sound change for voiced plosives or fricatives put next to unvoiced ones is that one will affect the other's voicing. But are some fricatives more likely to become stops when voiced than just be an unvoiced fricative?
I guess I'm just wondering if there are sound changes that are likely to affect certain phones in different ways than the rest of a set they might be in. Or if these changes that seem out of place are more random and it would be up to my instincts to throw them in when they feel right.
Also, for people who used to have a bad grasp on sound changes, what helped you get them a bit better? Do you stick to certain rules? Would you throw in changes like the one above or go for the expected β ð ɣ → f θ x / _{p,t,k,s} ?
Sound changes boggle me
Re: Sound changes boggle me
I don't know the specific language you're talking about, and some languages do just have weird sound changes.
However, what would generally make that rule make sense in most languages is the fact that many languages don't have a phoneme /x/.
When a rule acts on a phoneme, like devoicing, it results at least temporarily in a new bunch of potential phonemes. But there is pressure on a language to not continually increase the number of phonemes! So the new phonemes are pushed, as it were, to merge with existing phonemes.
This gives many potential sources of irregularity. Firstly, what the new phonemes merge with (and whether they merge at all) will depend on what phonemes the language already has.
In this case, when you devoice /B D G/, your immediate results are /P T x/ (P isn't the right symbol for a bilabial voiceless fricative, is it? Well, never mind, you know what I mean). Let's say your language ALREADY has /f T/, but no /x/. So, your language tries to have /P f/, but those phonemes are very close together; so /P/ can then merge with /f/. /T/ merges with existing /T/, no problem. So what happens to /x/? If the language doesn't already have /x/, then there are no other fricatives nearby for it to merge with, so it has to go further afield. One common option is for it to merge with /h/, also a backward fricative; the other common option is for it to merge with /k/, also a voiceless velar.
Remember, if your speakers don't distinguish /x/ as a phoneme, then new instances of [x] sound a lot like /k/ to them (/k/ may even have /x/ as an allophone already).
Here we have a second source of irregularity: which sounds are 'near' others? This is basically a question of which sounds tend to be distinguished from one another, and which tend to be merged. You can get a sense of this by looking at which sounds tend to exist in languages.
I don't have a simple set of rules. But here's some generalisations:
- fricatives tend to merge much more than stops do
- voiced fricatives and stops often merge
- /s/ is the 'strongest' fricative
- the further back you go in the mouth, the fewer sounds are likely to be distinguished
Regarding that last point: a bunch of languages have /p b t d k/ with no /g/. Lots of languages have /p t k f s/ with no /x/ (including English). In the case of English, [x] and [h] were once allophones; eventually [h] won out. Uvulars are even less likely to make all the distinctions than velars.
The other irregularities concern when a new phoneme merges away immediately, and when it sticks around. There's basically four factors here:
a) how 'strong' is the new sound and likely to remain distinguished?
b) how 'symmetrical' is the new sound system - does the new sound make the system logical and simple, or does it stick out as a weird exception to the overall pattern?
c) how many instances of the new sound are there? Is it a sound that only occurs in a few words, or is it going to be in half the vocabulary?
d) how important are the new instances? A sound that's found in 50 items of core vocabulary is more likely to survive than one that's found in 1000 super-rare words. In particular, the best way for a sound to survive as a phoneme is to occur in a commonly-used morphological affix.
-----------
This all assumes that what's really going on in your example is not /G/ > /k/ per se, but /G/ > /x/ > /k/ because /x/ merges right away into /k/.
If, on the other hand, your language does already have both /x/ and /k/, and yet the devoicing rule picks out /k/ rather than /x/, then that is indeed just weird.
[in that case, I'd probably ask whether /x/ was really [x]. For instance, if a language has a rule that /x/ is actually [h] before a consonant, then a new [x] before consonants might not be interpreted as /x/.]
However, what would generally make that rule make sense in most languages is the fact that many languages don't have a phoneme /x/.
When a rule acts on a phoneme, like devoicing, it results at least temporarily in a new bunch of potential phonemes. But there is pressure on a language to not continually increase the number of phonemes! So the new phonemes are pushed, as it were, to merge with existing phonemes.
This gives many potential sources of irregularity. Firstly, what the new phonemes merge with (and whether they merge at all) will depend on what phonemes the language already has.
In this case, when you devoice /B D G/, your immediate results are /P T x/ (P isn't the right symbol for a bilabial voiceless fricative, is it? Well, never mind, you know what I mean). Let's say your language ALREADY has /f T/, but no /x/. So, your language tries to have /P f/, but those phonemes are very close together; so /P/ can then merge with /f/. /T/ merges with existing /T/, no problem. So what happens to /x/? If the language doesn't already have /x/, then there are no other fricatives nearby for it to merge with, so it has to go further afield. One common option is for it to merge with /h/, also a backward fricative; the other common option is for it to merge with /k/, also a voiceless velar.
Remember, if your speakers don't distinguish /x/ as a phoneme, then new instances of [x] sound a lot like /k/ to them (/k/ may even have /x/ as an allophone already).
Here we have a second source of irregularity: which sounds are 'near' others? This is basically a question of which sounds tend to be distinguished from one another, and which tend to be merged. You can get a sense of this by looking at which sounds tend to exist in languages.
I don't have a simple set of rules. But here's some generalisations:
- fricatives tend to merge much more than stops do
- voiced fricatives and stops often merge
- /s/ is the 'strongest' fricative
- the further back you go in the mouth, the fewer sounds are likely to be distinguished
Regarding that last point: a bunch of languages have /p b t d k/ with no /g/. Lots of languages have /p t k f s/ with no /x/ (including English). In the case of English, [x] and [h] were once allophones; eventually [h] won out. Uvulars are even less likely to make all the distinctions than velars.
The other irregularities concern when a new phoneme merges away immediately, and when it sticks around. There's basically four factors here:
a) how 'strong' is the new sound and likely to remain distinguished?
b) how 'symmetrical' is the new sound system - does the new sound make the system logical and simple, or does it stick out as a weird exception to the overall pattern?
c) how many instances of the new sound are there? Is it a sound that only occurs in a few words, or is it going to be in half the vocabulary?
d) how important are the new instances? A sound that's found in 50 items of core vocabulary is more likely to survive than one that's found in 1000 super-rare words. In particular, the best way for a sound to survive as a phoneme is to occur in a commonly-used morphological affix.
-----------
This all assumes that what's really going on in your example is not /G/ > /k/ per se, but /G/ > /x/ > /k/ because /x/ merges right away into /k/.
If, on the other hand, your language does already have both /x/ and /k/, and yet the devoicing rule picks out /k/ rather than /x/, then that is indeed just weird.
[in that case, I'd probably ask whether /x/ was really [x]. For instance, if a language has a rule that /x/ is actually [h] before a consonant, then a new [x] before consonants might not be interpreted as /x/.]
-
LinguistCat
- sinic
- Posts: 325
- Joined: 06 May 2017 07:48
Re: Sound changes boggle me
Thanks Sal, this all makes a lot of sense. Hopefully I can keep a lot of it in mind for what I'm working on. It is a very clear way to think about the process, in any case.
I checked the context for that change, and while I couldn't find a starting phonology, it seems like nothing became /x/ and there was no other mention of /x/ before that change. But later some /k/ and some /ŋ/ both became /x/. So I think it was probably a case where [k~x] were /k/ for a while and then [x] split off as its own phoneme later, which I suppose is something else to keep in mind.
I checked the context for that change, and while I couldn't find a starting phonology, it seems like nothing became /x/ and there was no other mention of /x/ before that change. But later some /k/ and some /ŋ/ both became /x/. So I think it was probably a case where [k~x] were /k/ for a while and then [x] split off as its own phoneme later, which I suppose is something else to keep in mind.