(Ir)regularity of sound change

[Tropylium]

(Forked from the European languages before Indo-European thread.)

If they really did this, it is quite puzzling why sound changes still do run to completion, historically. And even more puzzling why you sometimes find what looks like a diffused sound change that in language A has been applied wider than in language B. We'd expect instead sound changes to leave leftovers all over the place, and to stop in all sorts of random places instead of nicely conditioned ones.

Sound changes often don't run to completion, historically.

"Often" is a statistical claim. It needs stronger evidence than anecdotal.

This is a big reason why there are irregularities in English spelling.

English features an exceptionally complex spectrum of sociolectal stratification and loaning between such registers, and is one of the worst possible places to be looking for examples of how sound changes happen in the first place.

I, for isntance, have a flat vowel in 'lather', even though other SSBE speakers have a broad vowel - despite the fact that this sound change happened centuries ago, and predates the trap-bath split that I do have.

And the very fact that you recognize that the sound change has already happened centuries ago, way before your idiolect even existed, is excellent evidence that everything you are talking about here amounts to dialectal and sociolectal reshuffling of shifted vs. unshifted cognates, and has nothing to do with the mechanism of the actual sound change.

Or take trans-, where we're seeing secondary analogisation: originally this prefix had broad A before a voiceless consonant, and flat before a vowel or voiced consonant, but this wasn't regular

"This change was regular with conditions foo, but it wasn't regular." Sure.

And the very concept of "irregular change" is epistemologically troubling. Clearly something must decide if a word in a given speaker's idiolect changes its shape or not. If it's not a change in the underlying phonology, then what? Are we to attach half-lifes to individual instances of phonemes? — And of course, does this not imply that "regular changes" consist of iterated irregular changes, and therefore the apparent regularity is merely an accidental epiphenomen of some sort?

No, because phonemes exist.

I fail to see how this addresses the criticism (or indeed, which part of it is this supposed to address at all).

The existence of probabilistic processes is not epistemically troubling at all.

No, but the idea that a process is probabilistic unless it can be proven regular is. We may lack the detailed dialectal and sociolinguistic records explaining where and how a word like Tausend came from; this does not mean it therefore "happened irregularly" (though I accept this as a descriptive shorthand), and more importantly it does not mean that the entire rest of the High German sound shift also originally happened irregularly.

Assumptions of in-progress irregularity are a problem for chain shifts as well. Sound change has no memory. If we assume that e.g. the Great Vowel Shift originally rampaged thru English words one by one, should we not find abundant examples where a word underwent one change such as /aː/ > /ɛː/ during an early period, then got hit by a wave of /ɛː/ > /eː/, slightly later by /eː/ > /iː/, and finally was targetted by a late torrent of /iː/ > /əi/? Nothing such happens at all.

That assumes that both the shift and the phonemic merger happen before the second part of the shift. In reality, it's not that /a:/ suddenly 'became' /E:/, it's that /a:/ moved toward the position where /E:/ was - that needn't mean that any lagging /E:/ words would necessarily merge with vanguard /a:/ words! Although I strongly suspect that you could find a few examples of this sort of thing if you looked.

I don't doubt you can find one or two examples. The key point is that this is not abundant at all, which is counterevidence for the claim that there ever existed a gradual shift of words from /aː/ to /ɛː/ one by one, without regard for regularity.

Regularity can of course be greatly mixed up afterwards by dialectal or sociolectal mixture. But such complications do not seem like sufficient evidence to conclude that all the historically known examples of regular conditioned sound changes happened as unsystematic "double-slit experiment" cascades of a word here, a word there, with no guiding overall principle observable during any one change.

Strawman!

It can only possibly be a strawman if I make claims about someone else's position. You know this, right? I am only working out the corollaries of Zomp's claim that "ongoing sound changes (…) diffuse word by word".

Most of the above mess around /a/ can be explained pretty simply:
- more commonly used words were more likely to broaden

i.e. at one point, there was a regular shift, in a basilect that lacked these less commonly used words.

- words that were perceived as more elevated, as words of foreign ancestry, or as more technical words were less likely to broaden

i.e. they were loaned from a non-shifted variety to a shifted one.

- words were more likely to broaden if they were semantically or phonologically associated with words that broadened; likewise, they were less likely to broaden (or more likely to un-broaden) if they were associated with words with flat vowels.

i.e. contamination is a process that exists.

Unfortunately, both differentiation by context and change through analogy are inherently particular processes - we can make generalisations, but not absolute rules.

I have no objection to this; I simply do not recognize either of these processes as "sound change" (and I think calling them such is misleading, much like etymological doublets due to loanwords also do not amount to sound change). They can only operate once a sound change has already happened.

[Salmoneus]

I'm not sure how to reply. The answer, after all, is very simple: the things you're saying are simply and demonstrably not true. You are taking an ideological position and ignoring the evidence of reality when it contradicts you - I don't see what fruitful discussion can come from that. If you're willing to throw out everything we know scientifically about languages (in this case, the history of English in particular), what are we meant to do to convince you?

It is simply not true that these sound changes happened regularly and then were 'contaminated' by other processes (processes that changed sounds but were not apparently sound changes, and processes that happen in every instance of sound change but must apparently be ignored as not 'real' sound change, because you insist on making your thesis unfalsifiable!). There is simply no reason to think this at all - and if it did happen, there would after all be a wealth of evidence (residual dialectical differences, comments by contemporaries on register differences, ill-educated spellings showing your alleged pronunciations, all the things that do show up for genuine sound changes). So essentially we're left with "X must always happen in all cases (except where it doesn't and that doesn't count), because I said so, and nothing can possibly change my mind about this". That's not a fruitful position to have a debate from.

[Pole, the]

I'm not sure how to reply. The answer, after all, is very simple: the things you're saying are simply and demonstrably not true. You are taking an ideological position and ignoring the evidence of reality when it contradicts you - I don't see what fruitful discussion can come from that.

Maybe he belongs to the ten percent?

[Basilius]

Suddenly, a quote from Labov.

I have examined the raising of /eyC/ as a prototypical sound change in a continuous phonetic space, below the level of conscious awareness, which has continued in the same direction for over a century. As an ideal candidate for a Neogrammarian sound change, it has been selected to test the diffusionist hypothesis that change propagates gradually through the vocabulary. The extended time period, the volume of data and the linear incrementation of /eyC/ has facilitated the comparison of the rates of change of individual words at different stages. Only a few candidates for lexical effects were detected in the course of this study, but none were decisive. A reasonable inference is that all words in which the phoneme occurs in the phonologically defined environment were selected simultaneously to participate in the change.

Given this finding, there remains the question of how typical this prototypical case may be. We know that lexical diffusion does occur, sometimes rooted in complex histories of dialect contact. The Philadelphia tensing of short-a has been repeatedly cited as a prototypical case of such lexical diffusion (Ferguson 1975, Labov 1989, 2001, Kiparsky 1988, Bermudez­‐Otero 2007). More recently, Fruehwald 2013 has advanced our understanding of the origins of lexical diffusion in the Philadelphia raising of /ay0/ before voiceless obstruents. But these are minority cases. So far, the study of a century of sound change in Philadelphia has found no evidence of lexical irregularity in the fronting of /aw/, /ow/ and /uw/, the raising of /ahr/ and /ohr/, the raising of /oh/or the backing of /e/, as well as the raising of /eyC/ examined here.

In the broader and less extended studies of the Atlas of North American English the great majority of the sound changes examined were found to be regular and lexical effects to be minimal (see Figure 1; Labov 2010, 10, Ch. 15).

(Underlining added by me.)

[zompist]

In Principles of Linguistic Change: Internal Factors, Labov suggests that some sound changes are regular, while others proceed by lexical diffusion (p. 542).

At the same time, his many diagrams (e.g. pp. 193, 205, 212, 216) label individual words; it's very clear that there are leading and trailing words in many of the ongoing sound changes he's studied.

I am only working out the corollaries of Zomp's claim that "ongoing sound changes (…) diffuse word by word".

Building straw men out of someone's second or third-hand report is not how you refute something. I'm not going to argue with your guesses and misconceptions, at least until you show that you've actually read some of the original authors and examined their evidence.

[Terra]

In French, words borrowed from Germanic that start with /w/ usually got changed to /g/ (via /gw/), but some native words from Latin got caught in the change too:
gaine < vagína
gâter < vastáre
guivre < vípera
Compare the normal development:
vouloir < valére

Also, Vulgar Latin had a sporadic change of k -> g:
Sp gritar < quiritāre
Sp golpear < VL colŭpus < colăphus

I know that there's others, but I can't recall them atm.