Sound Change Quickie Thread

[Pabappa]

well Greek did phth khth > /ft xt/, but I cant be sure what the path was. It could be that the second element of the series never lenited in the first place ... this would be similar to what happened in Germanic a little earlier on.

Which is why I quoted the Greek developments from /euθ/ and /eus/ - the change from voiceless aspirates to fricative seems to be older than the conversion of the second element of the diphthong from vowel to fricative.

Yeah I saw that, but wasnt sure I was reading it right .... is there a typo in that post or did they really go to /pt/ and /ps/ ?

[KathTheDragon]

Germanic may have fricativised both voiceless stops in a cluster. That makes it easier to lose the dental in the development of seven from *septm.

Except it's simply impossible, due to many many other words showing that the second stop in a cluster remained a stop, unconditionally. The *t has to be lost by other means, and note that a final *t is needed to protect the final nasal from being lost, so either metathesis (*septm̥ > *sepm̥t) or analogical spread from 10 -> 9 followed by dissimilation (*septm̥ -> *septm̥t > *sepm̥t) are plausible scenarios.

[Knit Tie]

lateral affricates

Oooh?

[Knit Tie]

Also, how realistic is English-style alveolar approximant becoming a /z/?

[StrangerCoug]

Also, how realistic is English-style alveolar approximant becoming a /z/?

I buy it as a conditional fortition.

[Richard W]

well Greek did phth khth > /ft xt/, but I cant be sure what the path was. It could be that the second element of the series never lenited in the first place ... this would be similar to what happened in Germanic a little earlier on.

Which is why I quoted the Greek developments from /euθ/ and /eus/ - the change from voiceless aspirates to fricative seems to be older than the conversion of the second element of the diphthong from vowel to fricative.

Yeah I saw that, but wasnt sure I was reading it right .... is there a typo in that post or did they really go to /pt/ and /ps/ ?

To /ft/ and /ps/! For the former, see Greek λευτεριά. The second change can be seen in the development of Classical Greek ἔκαυσε(ν) to Modern Greek έκαψε 'he burnt'.

[Richard W]

Germanic may have fricativised both voiceless stops in a cluster. That makes it easier to lose the dental in the development of seven from *septm.

Except it's simply impossible, due to many many other words showing that the second stop in a cluster remained a stop, unconditionally. The *t has to be lost by other means, and note that a final *t is needed to protect the final nasal from being lost, so either metathesis (*septm̥ > *sepm̥t) or analogical spread from 10 -> 9 followed by dissimilation (*septm̥ -> *septm̥t > *sepm̥t) are plausible scenarios.

How do these other words show that fricativisation wasn’t followed by dissimilation? That's two simple rules instead of one complex rule, which may therefore even be a simpler hypothesis.

[KathTheDragon]

Germanic may have fricativised both voiceless stops in a cluster. That makes it easier to lose the dental in the development of seven from *septm.

Except it's simply impossible, due to many many other words showing that the second stop in a cluster remained a stop, unconditionally. The *t has to be lost by other means, and note that a final *t is needed to protect the final nasal from being lost, so either metathesis (*septm̥ > *sepm̥t) or analogical spread from 10 -> 9 followed by dissimilation (*septm̥ -> *septm̥t > *sepm̥t) are plausible scenarios.

How do these other words show that fricativisation wasn’t followed by dissimilation? That's two simple rules instead of one complex rule, which may therefore even be a simpler hypothesis.

The argument's quite simple: there's no corroborating evidence for your hypothesis beyond the word under discussion, so it's methodologically wrong to reject the Germanic spirant law in favour of your idea, whereas the traditional approach has just "seven" requiring a special explanation, which are forthcoming and not particularly awkward in the slightest.

[Knit Tie]

Can a language lose its affricates, and then redevelop dʒ and tʃ and start treating them as consonant sequences?

[Whimemsz]

Sure. As already noted, speakers are blind to previous sound changes, so whatever happened before is irrelevant. So all you're asking is "can a language develop dʒ and tʃ and treat them as consonant sequences rather than affricates?" -- to which the answer is yet.

[Richard W]

The argument's quite simple: there's no corroborating evidence for your hypothesis beyond the word under discussion, so it's methodologically wrong to reject the Germanic spirant law in favour of your idea, whereas the traditional approach has just "seven" requiring a special explanation, which are forthcoming and not particularly awkward in the slightest.

It seems that a key part of the 'Germanic spirant law' is that Grimm and Verner did not apply to sequences of obstruents. Have I got that right? It looks more like a description of the result than a specific change, and I see that there's some unease about it in the wikipedia talk page. It also seems touched by issues related to Lachmann's and Bartholomae's laws, namely sound changes infecting phonological processes. It certainly reinforces the notion that /s/ behaves differently when it comes to manner assimilation or dissimilation.

[KathTheDragon]

It seems that a key part of the 'Germanic spirant law' is that Grimm and Verner did not apply to sequences of obstruents. Have I got that right? It looks more like a description of the result than a specific change

The spirant law is best explained as an exception to Grimm's law - after an obstruent voiceless stops do not become voiceless fricatives as they normally would. It's less clear how Verner's law factors in, principally because the timing (and hence the exact description) of Verner's law isn't clear.

and I see that there's some unease about it in the wikipedia talk page.

Looking at the talk page, I'm not quite sure where the confusion comes from. CodeCat says

It is in fact highly doubtful that the original consonant of the preterite was -t- especially considering its relationship to 'do'

but this is in fact not the general assumption in the literature - the consonantism is assumed to be the *t of the past participle, and so all the voiceless clusters are entirely expected. Imo Bartholomae's law in Indo-Iranian proves nothing about voicing assimilation within PIE, as I know of no examples that can only be explained as an inherited mixed-voicing cluster, rather than an analogically restored cluster within the history of PIIr.

It also seems touched by issues related to Lachmann's and Bartholomae's laws, namely sound changes infecting phonological processes.

Elaborate?

(Aside to mods, perhaps this discussion should be split into its own thread?)

[Richard W]

It seems that a key part of the 'Germanic spirant law' is that Grimm and Verner did not apply to sequences of obstruents. Have I got that right? It looks more like a description of the result than a specific change

The spirant law is best explained as an exception to Grimm's law - after an obstruent voiceless stops do not become voiceless fricatives as they normally would. It's less clear how Verner's law factors in, principally because the timing (and hence the exact description) of Verner's law isn't clear.

and I see that there's some unease about it in the wikipedia talk page.

Looking at the talk page, I'm not quite sure where the confusion comes from. CodeCat says

It is in fact highly doubtful that the original consonant of the preterite was -t- especially considering its relationship to 'do'

but this is in fact not the general assumption in the literature - the consonantism is assumed to be the *t of the past participle, and so all the voiceless clusters are entirely expected. Imo Bartholomae's law in Indo-Iranian proves nothing about voicing assimilation within PIE, as I know of no examples that can only be explained as an inherited mixed-voicing cluster, rather than an analogically restored cluster within the history of PIIr.

It's Blaschke's comments which express the unease. Taking the bold approach on PIE assimilation, the developments of PIE *tt (arguably [t͜st], i.e. with an affricate allophone of *t - that also works nicely with *tk), *st, *ps, *ts and *ks may not even belong under this law, and it seems that the German name given there is something different, but with some overlap.

Now, going back to the more widespread explanation of an exception to Grimm's law, why is having exceptions methodologically sounder than having a fix-up laws that then bring about the exceptions? Now, it does make me wonder whether *sk > *sx (Grimm) > *sk (fix-up/exception) wasn't complete throughout Germanic - though 'presigmatised stops' is a whole other topic. Am I missing a point about the 'Germanic spirant law' capturing a 'conspiracy', i.e. summing up several laws that together bring about a particular result as though by design?

It also seems touched by issues related to Lachmann's and Bartholomae's laws, namely sound changes infecting phonological processes.

Elaborate?

The analogical restorations implied by their application to past participles might not have been audible until Lachmann's law (Latin, applying to an assimilatory devoiced morpheme-final voiced consonant) and Barthomolae's law (Indo-Iranian) started to take effect. This is an unsettling idea for anyone who believes that sound laws apply to surface forms. The Wikipedia author seems to have tried to play safe by assuming that the voicing assimilation didn't necessarily happen within PIE.

(Aside to mods, perhaps this discussion should be split into its own thread?)

I concur.

[KathTheDragon]

It's Blaschke's comments which express the unease. Taking the bold approach on PIE assimilation, the developments of PIE *tt (arguably [t͜st], i.e. with an affricate allophone of *t - that also works nicely with *tk), *st, *ps, *ts and *ks may not even belong under this law, and it seems that the German name given there is something different, but with some overlap.

*st is obviously descriptively part of the law, as it is an exception to Grimm's. I have no idea why anybody would include *Ts since there is no exception. I'm not familiar with whatever the German word is referencing.

Now, going back to the more widespread explanation of an exception to Grimm's law, why is having exceptions methodologically sounder than having a fix-up laws that then bring about the exceptions? Now, it does make me wonder whether *sk > *sx (Grimm) > *sk (fix-up/exception) wasn't complete throughout Germanic - though 'presigmatised stops' is a whole other topic. Am I missing a point about the 'Germanic spirant law' capturing a 'conspiracy', i.e. summing up several laws that together bring about a particular result as though by design?

The simple answer is that while there's no reason it couldn't be a second sound law, there's also no reason to assume it was. It doesn't solve any additional questions, there's no independent evidence for it being separate. It's tidier to just list it straightforwardly as a conditioning environment of Grimm's law.

It also seems touched by issues related to Lachmann's and Bartholomae's laws, namely sound changes infecting phonological processes.

Elaborate?

The analogical restorations implied by their application to past participles might not have been audible until Lachmann's law (Latin, applying to an assimilatory devoiced morpheme-final voiced consonant) and Barthomolae's law (Indo-Iranian) started to take effect. This is an unsettling idea for anyone who believes that sound laws apply to surface forms. The Wikipedia author seems to have tried to play safe by assuming that the voicing assimilation didn't necessarily happen within PIE.

This has been solved to my satisfaction. There is no need to assume that PIE tolerated mixed-voicing clusters.

[Richard W]

The simple answer is that while there's no reason it couldn't be a second sound law, there's also no reason to assume it was. It doesn't solve any additional questions, there's no independent evidence for it being separate. It's tidier to just list it straightforwardly as a conditioning environment of Grimm's law.

Does that work out with sophisticated Sound Change Appliers? With the simple ones I've used, adding exceptions is usually a major complication.

This has been solved to my satisfaction. There is no need to assume that PIE tolerated mixed-voicing clusters.

I think the example of Ukrainian dropping assimilation is very significant. Of course, that does mean that one has less idea of what Germanic had in the way of clusters at morpheme boundaries at the time of Grimm's law.

[Man in Space]

You can easily add exceptions in phonix, for what it’s worth.

[bradrn]

Zompist’s SCA² lets you add exceptions pretty easily as well. So does my own SCA.

[KathTheDragon]

The simple answer is that while there's no reason it couldn't be a second sound law, there's also no reason to assume it was. It doesn't solve any additional questions, there's no independent evidence for it being separate. It's tidier to just list it straightforwardly as a conditioning environment of Grimm's law.

Does that work out with sophisticated Sound Change Appliers? With the simple ones I've used, adding exceptions is usually a major complication.

My SCA can also do exceptions. It's really not that complicated, actually.

Also, why are you even asking about SCAs?

This has been solved to my satisfaction. There is no need to assume that PIE tolerated mixed-voicing clusters.

I think the example of Ukrainian dropping assimilation is very significant. Of course, that does mean that one has less idea of what Germanic had in the way of clusters at morpheme boundaries at the time of Grimm's law.

Maybe so, but as I'm sure you can guess by this point, there's no evidence that Germanic did much in the way of restoring root-final consonants in assimilating contexts.

[Nortaneous]

lateral affricates

Oooh?

I've met native speakers of English with lateral affricates instead of postalveolars; given the Turkmen lisp and the Armenian and Polish l-dealveolarization, sure, why not

[Richard W]

The simple answer is that while there's no reason it couldn't be a second sound law, there's also no reason to assume it was. It doesn't solve any additional questions, there's no independent evidence for it being separate. It's tidier to just list it straightforwardly as a conditioning environment of Grimm's law.

Does that work out with sophisticated Sound Change Appliers? With the simple ones I've used, adding exceptions is usually a major complication.

My SCA can also do exceptions. It's really not that complicated, actually.

Also, why are you even asking about SCAs?

Measures of tidiness. Also from the point of view of having a pre-packaged set of sound laws to choose from, what should "Grimm's Law" look like?