zompist bboard

All of you who claim to have two schwas, a high one closer to the PIT vowel and a low one more closer to the STRUT vowel, i.e. the Rosa's-roses distinction - should take a read of this language log post. While the post is aimed at a slightly different central point, most of the material is directly relevant to the two-schwas question.

Of particular interest is this passage:

Mark Liberman wrote:Many people, including some famous phoneticians, have advanced a contrary view, namely that (at least in some varieties of English) there are two distinct schwas, one higher and fronter, the other lower and backer. A supposed near-minimal-pair, for some of the schwa-splitters of my acquaintance, is Alice vs. Dallas. I've never seen any evidence from natural speech that supports this view — performance of (near-) minimal pairs is always suspect in such cases, since it's subject to facultative disambiguation.

There's further discussion in the rest of the post, and in the comments. The broader lesson of which is that our phonetic intuitions about it have little apparent relationship to the available evidence.

...yeah, no.

While it's all fine and dandy busting out the skepticism when it comes to marginal distinction (eg, do I have one schwa, or are many of my schwas just strongly reduced yet distinct versions of the different unreduced vowels (in particular, do I have a rounded schwa?)?; or do I ever assimilate schwas to following vowels as the LL entry discusses), but that doesn't mean that we can seriously claim that our 'phonetic intuitions' can't be trusted in clear-cut cases, like /k/ vs /m/ or, in this case, /@/ vs /I/.

If you're going to say that "Rosa's" and "roses" have the same second vowel, why not say that the first vowel is the same as well? That would make just about as much sense!

As far as I'm concerned, the second vowel in roses is not a shame at all.

Salmoneus wrote:While it's all fine and dandy busting out the skepticism when it comes to marginal distinction ... , but that doesn't mean that we can seriously claim that our 'phonetic intuitions' can't be trusted in clear-cut cases, like /k/ vs /m/ or, in this case, /@/ vs /I/.

"Evidence is irrelevant if I believe really really hard in my preferred answer!" is an even less valid argument than KathAveara's word salad, and that is truly impressive.

Salmoneus wrote:If you're going to say that "Rosa's" and "roses" have the same second vowel, why not say that the first vowel is the same as well? That would make just about as much sense!

it makes perfect sense to say that, because they are

I hear ['ɻʷoʊ.zʌz 'ɻʷoʊzɪz] when I say them.

But like the article says, it could just be distinction when they're stressed.

Abi wrote:I hear ['ɻʷoʊ.zʌz 'ɻʷoʊzɪz] when I say them.

You said the words while consciously listening for differences, and perhaps mid-consciously producing said differences. I run into this issue so often that I've given a pet name to it: the gremlin. A little gremlin that gets in your head and mouth and messes with your ability to pronounce things as you would normally would, cause you're thinking too hard about it, and it makes self-diagnosis impossible. You can't really know how you normally say those words unless someone else manages to record you saying them in a situation where you aren't thinking about it.

This is like the <jeune> vs <jeûne> issue in French. Pretty much all grammars, but many linguists as well, argue that these two words constitute a minimal pair: <jeune> /ʒœn/ (young) vs <jeûne> /ʒøn/ (fasting). And in fact this minimal pair is the *only* minimal pair for a /ø/~/œ/ contrast in French. And when you make people read these two words side by side, in isolation, recordings show that they are indeed able to make that contrast.


The problem is that when you make them read these words *in context*, distant from each others, the contrast disappears, and both words can be read either as [ʒøn] or [ʒœn]. There is only really one phoneme /ø/ of which [œ] is an allophone that exists in free variation in many environments.


This implies that minimal pairs can be artificially deduced when there are none if care is not taken to erase the bias of people's prejudice about their own language (French people are *explicitely taught* that they should pronounce <jeune> and <jeûne> distinctly; I suspect something similar is going on for <rosa's> and <roses>).

Bugger, I forgot that autocorrect was on.

KathAveara wrote:As far as I'm concerned, the second vowel in roses is not a shame at all.

I meant to say that it's not a schwa. Hopefully that makes more sense.

Radius Solis wrote:

Salmoneus wrote:While it's all fine and dandy busting out the skepticism when it comes to marginal distinction ... , but that doesn't mean that we can seriously claim that our 'phonetic intuitions' can't be trusted in clear-cut cases, like /k/ vs /m/ or, in this case, /@/ vs /I/.

"Evidence is irrelevant if I believe really really hard in my preferred answer!" is an even less valid argument than KathAveara's word salad, and that is truly impressive.

The problem is that Liberman's position is the same. He's really really sure that there's no difference, and that anyone who hears one is just mistaken. From the comments, it seems that Brits are the most insistent about the double shwas. (Liberman is American.) It would seem that's a good place to investigate more thoroughly.

I do take the warning that we can easily confuse ourselves. From looking at linguists' attempts to transcribe the same material, I suspect that getting vowels one step off is just routine.

I've been going through my English pronunciation file testing out [ə] and [ɪ], with the result that I don't know what I'm hearing anymore. A word like accusative shows off both very distinctly; on the other hand you could argue that there's secondary or tertiary stress on the -ive. Conscious / accomplish contrast, but on the other hand I'm beginning to suspect an allophonic rule: [ɪ] seems more likely in a closed syllable but not before [m n l r s]. (Except for [s], the contrast is not with [ə] but with a syllabic consonant.) In an open syllable I think I have a weak contrast in phot[ə]graph / eluc[ɪ]date, but I wouldn't be surprised to be shown that I'm inconsistent there.

Marian Macchi's observation on shwa anticipating the next vowel is very interesting.

http://languagelog.ldc.upenn.edu/nll/?p ... ent-395611

Edward Flemming wrote:A paper I wrote with Stephanie Johnson a few years ago has data on the Rosa's-roses distinction in American English (it is a distinction for most speakers.)

The summary is italicized.
(the relevant paper is here.)

Of course this is still very much a good tendency to keep in mind when thinking about phonetics/linguistics (that distinctions not actually present, such as jeune/jeûne, can still be perceived as existing by (native) speakers.)

Legion wrote:This is like the <jeune> vs <jeûne> issue in French. Pretty much all grammars, but many linguists as well, argue that these two words constitute a minimal pair: <jeune> /ʒœn/ (young) vs <jeûne> /ʒøn/ (fasting). And in fact this minimal pair is the *only* minimal pair for a /ø/~/œ/ contrast in French. And when you make people read these two words side by side, in isolation, recordings show that they are indeed able to make that contrast.

The problem is that when you make them read these words *in context*, distant from each others, the contrast disappears, and both words can be read either as [ʒøn] or [ʒœn]. There is only really one phoneme /ø/ of which [œ] is an allophone that exists in free variation in many environments.

This implies that minimal pairs can be artificially deduced when there are none if care is not taken to erase the bias of people's prejudice about their own language (French people are *explicitely taught* that they should pronounce <jeune> and <jeûne> distinctly; I suspect something similar is going on for <rosa's> and <roses>).

I found Liberman's claim that he can hardly distinguish ladder from latter weird for the same reason. I've pointed out this pair of words to many speakers here, and every time I've done it they pronounce latter as [ˈlætʰɹ̩] when said next to ladder [ˈlæɾɹ̩] (however, they often also self-doubt about the [tʰ] bit, pronouncing latter [ˈlæɾɹ̩] and [ˈlætʰɹ̩] interchangeably).

Radius Solis wrote:

Salmoneus wrote:While it's all fine and dandy busting out the skepticism when it comes to marginal distinction ... , but that doesn't mean that we can seriously claim that our 'phonetic intuitions' can't be trusted in clear-cut cases, like /k/ vs /m/ or, in this case, /@/ vs /I/.

"Evidence is irrelevant if I believe really really hard in my preferred answer!" is an even less valid argument than KathAveara's word salad, and that is truly impressive.

Congratulations, you have discovered skepticism. Yes, everything we think we know MIGHT be wrong. Similarly, there's no point doing, say, spectrographic analysis, because while you might THINK that the graphs say one thing, you MIGHT be wrong. You might be deluding yourself, or being mislead by a malicious demon, or you might be living in a virtual reality constructed by robots determined to make you think that /I/ is not a phoneme.

But the fact that you can IMAGINE commonsense impressions being misleading is not evidence that they ARE misleading.

You've taken two random phonemes and asserted that they're the same. And your evidence for this is that... everyone might be wrong about being able to distinguish them. Well yes, they might, and I guess that's a good reason not to stake the existence of the universe on the assumption that what seems true is true. But that's not actually evidence that they aren't distinguished.

This would be more appealing if you'd chosen some apparent 'fact' that seemed dubious or marginal, the sort of thing small enough for people to be easily mislead about it. People thinking that the stop in 'pin' is the same as the stop in 'spin', for instance - that's the sort of thing where the more you think about it, the less clearly true it appears, and eventually you can actually adduce evidence that they're not (i.e. putting your hand in front of your mouth). But 'do "Rosa's" and "roses" have the same vowel in the second syllable?' is such an obvious and unambiguous 'yes' that it would take some really persuasive evidence to overcome the evidence of the senses. Given that it's an issue where popular opinion, careful introspective evidence and the opinions of experts all coincide... sorry, but it'll take more than reasoning of the 'ahh, but how do you KNOW your teeth aren't made of cheese!?' kind to make the contrary opinion seem respectable. Because yes, popular opinion, careful introspective evidence and the opinions of experts CAN all be wrong, and can even all be wrong at the same time... but that's not a reason to believe that they ARE wrong!

Otherwise you float down the Pyrrhonian/Ungerian "oh but nobody can ever know anything because any belief that there is evidence may itself be a mistaken belief" waterway to sitting very still not able to do anything in case the imaginary crocodile eats you, because you can never know for sure that it's NOT there.

ETA: maybe you're confused because "Rosa's" and "roses" aren't words that often have to be distinguished. But there can surely be no reasonable doubt that the vowels are different in 'Lennin' and 'Lennon', 'purest' and 'purist', or 'illusion' or 'allusion'!

Another minimal pair: razors vs raises.

P.S. if you're determined to do away with schwa, it seems a lot more sensible to say that non-reduced schwa is an allophone of /r/. Though this isn't entirely convincing itself.

ITT: Salmoneus forgets everyone doesn't speak his dialect.

Salmoneus wrote:ETA: maybe you're confused because "Rosa's" and "roses" aren't words that often have to be distinguished. But there can surely be no reasonable doubt that the vowels are different in 'Lennin' and 'Lennon', 'purest' and 'purist', or 'illusion' or 'allusion'!

First off, I'd use something besides ETA, since for most people I'd think that would be interpreted as estimated time of arrival long before they realized it might be edited to add (I didn't find it until my 5th site while searching.)
Secondly, in my version of General American, your three sets of example words are homophonous (3 pronunciations, for the record).

Another minimal pair: razors vs raises.

Someone's forgotten rhotic dialects exist.

P.S. if you're determined to do away with schwa, it seems a lot more sensible to say that non-reduced schwa is an allophone of /r/. Though this isn't entirely convincing itself.

Since no one is determined to do away with schwa, that's not really a problem.

ITT: Drydic forgets that Radius forgets that not everyone speaks his dialect.

Radius wasn't making a claim about GA specifically, but about English in general - about all English. Hence beginning with "for all of you who claim to have... the Rosa's-roses distinction". That includes me, so talking about my dialect is entirely relevant.

And this is functionally the same as doing away with schwa as a phoneme - saying that schwa does not have minimal pairs with the one other vowel phoneme that minimal pairs exist for is basically saying that schwa is not phonemic. If schwa is the same as /I/, then /@/ is not a distinct phoneme.

-----------

Radius: sorry if my tone there was a tad abrasive. It's just that what with having a philosophy degree, I've seen a lot of this "but how can you be sure you can trust the 'evidence' of your senses, given that we know that the senses can sometimes be mistaken or misinterpreted?" business - hell, I went through that phase myself. But sooner or later you have to buckle down and deal with the real world, in which the possibility of error is not evidence against any position. It's not even a reason to discount the potentially erroneous evidence. Eyewitness testimony in court, for instance, is known to be extremely susceptible to error, yet it is still evidence - still relatively good evidence, even. More generally, the existence of optical illusions and delusions of memory does not mean that vision and memory are not sources of evidence - good and realiable sources of evidence, even. Likewise, the fact that in some cases introspective speech analysis can be mislead either directly by perceptual bias or indirectly through assumption-lead altered production is not a reason why introspective speech analysis cannot be regarded as good evidence. It's just a reason why it shouldn't be regarded as indisputable evidence, when contradictory evidence arises. But we're talking here about a case where you have no contradictory evidence - in which case the prima facie evidence from introspection ought to be allowed to stand. Likewise, it is reasonable to believe that Liverpudlian English contrasts /E/ and /p/, even though I doubt any specific studies have been done on the question...



While we're on the topic, however, I'd also like to make a second point, which is that I'm skeptical of this widespread idea that what matters is only slurred and mumbling speech - that if in "real" speech a distinction can't be detected, the distinction isn't really there, even if it appears when people speak slowly or carefully, or when they are intentionally disambiguating. [For instance, I might not always distinguish Rosa's and roses in casual speech - not because both are /@/, but because I probably drop the second vowel entirely in some cases and just say [roUz.z]. But then when you get to that level of casual speech, all hell breaks loose anyway.]
There are two ways of looking at this, it seems to me. One way is to say that basic form of every word is how that word appears in the most rapid, reduced, unclear utterances. In this case, each word has its own basic form, and then each word requires its own ad hoc rules telling you which sounds get added when speakers are speaking more clearly. Which is very complicated. Or there's the other way, which is to say that the clear pronunciations reflect underlying phonemes, which are progressively neutralised (and sometimes subject to liaisons and attractions and whatnot) in more and more careless speech. Which in most cases seems to me to be a far more parsimonious way of explaining things.

This post by John Wells may have some information as well.


JAL

Salmoneus wrote:But there can surely be no reasonable doubt that the vowels are different in 'Lennin' and 'Lennon', 'purest' and 'purist', or 'illusion' or 'allusion'!

Those pairs are each entirely homophonous for me.

http://www.imdb.com/title/tt0118715/quo ... =qt0464769

Couldn't schwa be an allophonic realization of either /6/ or /I/? That way, you can admit that it's not phonemic while maintaining the "two schwa" contrast.

You're talking about the British two schwas, right?

I for one definitely have the Rosa's-roses distinction, but as mentioned in the past it is actually a morpheme boundary distinction than the second vowel therein itself differing phonemically.

Salmoneus wrote:But there can surely be no reasonable doubt that the vowels are different in 'Lennin' and 'Lennon', 'purest' and 'purist', or 'illusion' or 'allusion'!

I distinguish the first and third (assuming Lennin, with which I am unfamiliar, is said the same as Lenin), but not the second. I have lower midde class RP vowels with a slight northern influence.

You are aware, I trust, that 'illusion' and 'allusion' are readily confused, and that 'accept' is frequently pronounced the same as 'except'.

Richard W wrote:

Salmoneus wrote:But there can surely be no reasonable doubt that the vowels are different in 'Lennin' and 'Lennon', 'purest' and 'purist', or 'illusion' or 'allusion'!

I distinguish the first and third (assuming Lennin, with which I am unfamiliar, is said the same as Lenin), but not the second. I have lower midde class RP vowels with a slight northern influence.

You are aware, I trust, that 'illusion' and 'allusion' are readily confused, and that 'accept' is frequently pronounced the same as 'except'.

Seconded. Also, I merge elusion with illusion, but keep allusion distinct. Mostly GenAm.

Yeah I have in <purest> and <purist>, and I often see confusion in spelling ('my specialist girly' is one example that springs to mind from The Internet)