American English : are [ə] and [ʌ] different phonemes? (schwa vs. chevron)

Question

What case can be made for considering whether [ə] and [ʌ] are different phonemes or not in American English? Please note the focus is on standard American English. EDIT: i.e.: on General American.

Many dictionaries use /ʌ/ in stressed position and /ə/ in unstressed positions. So we get transcriptions such as:

DUST /dʌst/
LOVE /lʌv/
BACKUP /ˈbækˌʌp/
KETCHUP /ˈkɛtʃ.əp/
CUSTOM /ˈkʌs·təm/

However, if these truly constitute two different phonemes, then we should be able to come up with minimal pairs to illustrate the contrast between both sounds.

I cannot think of a single minimal pair to contrast /ə/ and /ʌ/.

It's interesting to look at CMU Dictionary, the pronouncing dictionary of American English.

CMU Dictionary uses AH for both sounds. So we get:

DUST D AH1 S T
LOVE L AH1 V
BACKUP B AE1 K AH2 P
KETCHUP K EH1 CH AH0 P
CUSTOM K AH1 S T AH0 M

(0 = unstressed, 1 = primary stress, 2 = secondary stress)

My understanding is that, provided the stressed syllables are pronounced longer and with more energy, saying BACKUP, KETCHUP and CUSTOM as [ˈbækˌəp], [ˈkətʃ.əp] and [ˈkəs·təm] would not hinder comprehension in the least.

As I see it, there's only one phoneme here, which happens to be realized [ʌ] in stressed position and as a schwa [ə] in unstressed positions.

This leaves us with two problems:

Problem 1 : if /ə/ and /ʌ/ are different phonemes, what are some examples of minimal pairs between the two?
Problem 2 : if they are allophones, which notation should be used for the phoneme? I assume /ə/.

The reason I'm asking is because I'm teaching American English with a lot of phonemic transcriptions – why teach two phonemes when there's only one.

Finally, where would professional linguists locate those sounds on the following chart (again, from an American point of view)?

The only discussion of the topic I found online so far is on this forum. There's also, to some extent, the Talk section about the ARPAbet article on Wikipedia.

EDIT (DEC 2018): FYI, I was asking the question in the context of designing an IPA chart for American English (General American), to teach phonetics and ESL, which implies deciding just which phones to teach -- and why.

You have some wrong assumptions in your post. 1. "if these truly constitute two different phonemes, then we should be able to come up with minimal pairs" - Not necessarily. There are phonemes with limited distribution (e.g. h, ŋ). etc. 2. As Giegerich 1992 correctly observes, "taking stress into account, schwa is in complementary distribution with all other vowels (except [ɪ])" and thus, he concludes, we are not entitled to call schwa an English phoneme. — Alex B., Oct 01 '16 at 17:54
The Wikipedia article dedicated to English phonology states "[ʌ] (stressed) and [ə] (unstressed) may be considered allophones of a single phoneme in General American", citing John Wells' Accents of English pp. 121, 132 (but unclear in which of the three volumes.) If I can get my hands on the book, I'll post an update and quote what I've found. — Fabien Snauwaert, Apr 04 '19 at 16:19
A nice video, America, we need to talk about STRUT ʌ and schwa ə discussing the issue and the amount of convention and confusion around it. — Fabien Snauwaert, Aug 09 '22 at 14:59

score 16 · Answer 1 · answered Oct 01 '16 at 14:16

16

This is a well-written argument, but I think it's mistaken to conclude that they are the same phoneme; or, more to the point, I think this is a case that highlights a limit of phoneme/allophone analysis.

Indeed, the same argument can also be used to show that /ə/ and /ʊ/ are allophones of the same phoneme: there cannot be a minimal pair because of the restriction of the stress patterns on the syllables they occur in. But then there is a problem, since there are plenty of minimal pairs distinguishing /ʌ/ and /ʊ/ (e.g. put/putt).

(One possible solution to this problem, rather than introducing /ə/ as a phoneme, would be to argue that /ə/ is perceptually closer to /ʌ/ than to /ʊ/, which does not match my personal intuition but seems closer to what other native English speakers tend to think.)

answered Oct 01 '16 at 14:16

hunter

792
4
13

3

I don't see a problem here. Can't you just say that [ə] is the unstressed allophone of both /ʌ/ and /ʊ/? Neutralization is hardly an unusual phenomenon, especially in this context. – user54748 Oct 01 '16 at 22:11
2

@user54748 i guess the point is that "x and y are allophones of the same phoneme" sounds like an equivalence relation, but it's not, since x ~ y and y ~ z does not imply x ~ z. – hunter Oct 20 '16 at 21:19
1

As a non-native, I have trouble distinguishing /ə/ and /ʌ/ in the first place – not the canonical IPA unrounded [ɔ], that one's fine; but the way /ʌ/ is realized in American English, specifically, sounds to my barbarian ears super "the same thing"-y as [ə]. – melissa_boiko Jun 15 '18 at 07:31
1

@melboiko yes, quite. [ʌ] is not a good fit for English/American /ʌ/. It's rather a spelling/transcription convention. But for most speakers it seems to be lower than /ə/ and a bit retracted in America. – unknown_person_1000 Dec 20 '18 at 20:59

score 6 · Answer 2 · answered Oct 01 '16 at 15:38

[ə] and [ʌ] are allophones of a single phoneme. Schwa appears in an unstressed syllable and wedge appears in a stressed syllable. Because of this complementarity, it is not possible to find minimal pairs distinguishing the vowel. However, the analysis has to be performed on phonetic transcriptions and not an assumed phonemic reduction of the phonetics, thus the pair [ˌsʌbˈvɹ̩ʒn̩] (sub-version) and [səbˈvɹ̩ʒn̩] (subversion) is not a minimal pair, because the initial syllables differ in stress.

An alternative would be to claim that stress is phonetically predictable, but is sensitive to whether or not the vowel in question is schwa (schwa cannot be stressed). Minimal pairs like the noun / verb contrast "pervert" [ˈpɹ̩vɹ̩t] / [pɹ̩ˈvɹ̩t] establish that stress is not phonetically predictable in English (a point that is well-known).

The logic of phonemic analysis alone does not dictate that schwa and wedge must be reduced to a single phoneme, it says that they may. This is a classic problem of phonemic analysis discussed in the classic paper "The non-uniqueness of phonemic solutions of phonetic systems" by Y.R. Chao. If you add to the theory an Occam's Razor compulsion to minimize the size of the phonemic inventory (which was also done), then you must reduce these two phones to one phoneme. Since in phonemic analysis, phonemes are classifications of diverse sets of phones, it is a category error to think that phonemes have phonetic properties. It is however reasonable to wonder what would be a convenient way to write that phoneme: I personally like schwa, but there is nothing wrong with wedge as a letter. More languages use schwa as an official letter that use wedge, if that matters.

Your logic chart does replicate the flow of thought used in popular accounts of the concept phoneme. Phoneme choice does not "change meaning" (if you think about what "change" means, that should be clear), and failure to "change meaning" does not mean that the sounds are in free variation. The minimal pair test is an absolute test for phonemic status: if [a] and [b] appear in the same phonetic substitution frame ([bɪp] "horse" vs [bɪb] "arrow") then the sounds are instantly proven to be distinct phonemes. If there are no such pairs, it is still possible that they can be reduced to a single underlying phoneme. The crucial question is whether some set of rules can be posited to derive the phonetic outputs: if not, then the two sounds cannot be distinguished purely by rule, and there must be two phonemes.

Free variation, on the other hand, refers to the situation where the same word can be pronounced in two different ways, and there is no linguistic distinction between the pronunciations (one may find social distinctions, such as whether the pronunciation [ð] is in a more informal register compared to [d] (in Palauan). The key is that free variation is defined in terms of "same word", not "changing meaning". Different words, of course, often have different meanings, but not always (sofa, couch).

Just to be clear, by the same argument [ə] is also an allophone of most other English vowels, correct? If so, it seems to me we would not want to call the underlying phoneme /ə/. — TKR, Oct 08 '16 at 16:41
The standard understanding of "allophone" is that phonemes and allophones are in a biunique relation. Neutralization of unstressed vowels is not context-free, for example the unstressed initial vowel of "exact" is [ɛ], not [ə]. So there is a non-allophonic rule neutralizing short unstressed vowels in open syllables to something central (wedge is an underlying vowel of English), and then an allophonic rule saying that the vowel is [ə] if it is unstressed. "Underlying" and "phoneme" are not the same thing, just in case one has neutralization. — user6726, Oct 08 '16 at 16:52
I'm still missing something, or more than one thing. 1. By "biunique" do you mean that the same phone can't be an allophone of more than one phoneme? Isn't that simply neutralization (e.g. both /d/ and /t/ having [ɾ] as an allophone)? 2. If the non-allophonic rule only affects unstressed vowels in the first place, how can you have an allophonic rule "saying that the vowel is [ə] if it is unstressed"? (I also don't understand your last sentence, but maybe that's a topic for a separate question.) — TKR, Oct 08 '16 at 17:24
Yes, each phone translates to a single phoneme: that's the classical definition, and neutralization is the problem. Some people understand "phoneme" to mean "underlying sound", which is different, so this is a YMMV matter, and a reason why one should avoid unclear terms like "phoneme". As for 2, not every unstressed vowel becomes wedge as I explained; but every unstressed wedge becomes schwa. But it crucially has to be unstressed, so you can't say "every wedge becomes schwa". The allophonic rule does not only apply to the output of full-to-wedge reduction: they are separate rules. — user6726, Oct 08 '16 at 17:29
If that's the classical definition, it seems obviously untenable given that neutralization is a common phenomenon, so why not abandon it? On the (um) wedge issue, if wedge-to-schwa reduction doesn't apply to the output of full-to-wedge reduction, how do you get full-to-schwa reduction? And can you briefly explain the difference between "phoneme" and "underlying sound", or would that be better discussed as a separate question? — TKR, Oct 08 '16 at 17:37
Why would one retain the concept "phoneme" at all? We have the concept "segment": what does "phoneme" buy you? I didn't say that ʌ→ə doesn;t apply to the output of V→ʌ: that is why "only" is there. But these are separate questions. — user6726, Oct 08 '16 at 17:42
Oh, sorry, I missed "only". You can of course argue that "phoneme" is not a useful term, but if one does retain it (at least here, given that the question and your answer are couched in terms of phonemes and allophones), I can't see why it necessarily entails biuniqueness, unless merely by a definition that can easily be modified. — TKR, Oct 08 '16 at 17:49
@user6726 Who exactly (names of those phonologists please) thought that "each phone translates to a single phoneme"? — Alex B., Oct 09 '16 at 17:54
And most importantly, why should we care about such a narrow interpretation that belongs in a textbook on history of phonology? — Alex B., Oct 09 '16 at 18:21
You don't have to care about anything. You can just decide that /ʌ/ and /ə/ are different phonemes because you feel like they are. There are no correct answers, just socially popular answers. — user6726, Oct 09 '16 at 18:33

Rosie F · Answer 3 · 2016-10-01T11:00:41.730

4

For a minimal pair to contrast /ə/ and /ʌ/, how about: "subversion" meaning an act of subverting, and "subversion" as in version 1 subversion 1.2.

edited Oct 01 '16 at 11:00

answered Oct 01 '16 at 09:16

Rosie F

602
3
7

4

That's just a stress contrast. The point is whether stressed schwa contrasts with unstressed schwa, or whether stress defines complementary distribution. This is a theoretical argument, since phoneme and allophone are terms in particular theories -- and may well have different meanings in different theories. – jlawler Oct 01 '16 at 16:37
1

@jlawler, it is indeed "just" a stress contrast. But that's what the question is about -- whether there is a contrast between the unstressed phone [ə] and the stressed phone [ʌ]. Rosie's example shows that there is. It works fine for me. – Greg Lee Oct 02 '16 at 12:42
@Greg Lee, the stress contrast is in the first syllable 'sub', not the schwa[s] in 'version'. – amI Oct 12 '16 at 18:56
@amI, that's true. Did you think I implied something different? – Greg Lee Oct 12 '16 at 20:13
@Greg Lee, yes -- You said that the test is to find a contrast between the phones in question, so finding a stress contrast of a different phoneme in a different syllable does not satisfy the test. – amI Oct 14 '16 at 18:38
@amI, but the stress contrast is in the first syllable, in this example. Perhaps you don't understand the example. – Greg Lee Oct 14 '16 at 18:43
@Greg Lee - I goofed my comments. I meant that the shift of stress to the second syllable invalidates the test on the first syllable. A minimal pair must differ in only one place. (If you say 'subversion' with no stress anywhere then I think it is ambiguous.) – amI Oct 14 '16 at 18:50
@amI, in my understanding, there is no shift of stress to the second syllable. The primary stress is always on the second syllable, but the first syllable has a secondary stress, in one sense, but no stress, in the other sense. – Greg Lee Oct 14 '16 at 19:02
I see the main stress on the first syllable of 'sub-version'. I'll need Praat and an audio corpus to verify. – amI Oct 14 '16 at 19:13
@aml What do you mean by "place"? Do you consider the nature of the first syllable's stress as one "place" and the position of the vowel as another "place"? If so, then how do you respond to Greg's answer? – Rosie F Oct 14 '16 at 19:24

score 4 · Answer 4 · answered Jun 14 '18 at 23:21

4

Your question doesn't really have an answer.

For me, there is a contrast between the weak form of just meaning recently, /dʒəst/, and the word just meaning fair, /dʒʌst/.

I use the weak form of just almost all the time, and the vowel is definitely different from the one I use in just, meaning fair.

So I perceive them as different phonemes. But many Americans pronounce them exactly the same, and these Americans would probably perceive them as the same phoneme.

answered Jun 14 '18 at 23:21

Peter Shor

374
1
9

I completely understand that it is far more common to stress just 'fair' than just 'recently' but imagine you are annoyed that someone has repeatedly told you to close the window and you say: I've JUST done that! Here you have the strong form. – unknown_person_1000 Dec 20 '18 at 21:28
I never use a schwa in just "fair", even when it doesn't have much stress on it. And I use a schwa in just "recently" even when it has a moderate amount of stress on it. (Maybe not in I've JUST* done that,* but in phrases like hold on just* a minute.) And the two vowels are definitely different. In fact I can use either the strong form or the weak form in hold on just a minute*. The distinction between these is not the stress, but the vowels. – Peter Shor Dec 21 '18 at 16:18
I agree with this. Even when very emphatically stressed, just ‘recently’ and but both have schwa, never wedge; and even when unstressed, just ‘fair’ and butt always have schwa, never wedge. I cannot think of any way to reconcile that with them being a single phoneme. – Janus Bahs Jacquet Apr 30 '20 at 05:30

Greg Lee · Answer 5 · 2016-10-01T14:51:08.423

2

Stress is phonemic in English, so of course [ə] and [ʌ] are different phonemes, since only the latter is stressed.

edited Oct 01 '16 at 14:51

answered Oct 01 '16 at 13:14

Greg Lee

12,466
1
18
32

9

Why not just consider [ʌ] a stressed allophone of /ə/? Any contrast can be assigned to conditioning by stress, which is, as noted, phonemic, and is expected to condition vowels. – jlawler Oct 01 '16 at 15:00
3

That fact makes them more likely to be allophones, since it suggests that they are in complementary distribution. – anomaly Oct 01 '16 at 15:21
@anomaly, what is the "them" in your statement? I was talking about unstressed [ə] and stressed [ʌ]. Are these really in complementary distribution? – Greg Lee Oct 01 '16 at 15:52
4

@jlawler, There are several vowels which alternate with [ə]. You wouldn't be able to predict what vowel quality would result when /ə/ gets stressed. – Greg Lee Oct 01 '16 at 15:56
1

Many phonemes have several allophones in different environments, including stressed environments. There's no significant contrast among central vowels in English, unlike front and back. The result is that the entire central vowel space is effectively occupied by one phoneme whose allophones are conditioned by virtually everything -- initial/nucleic/final position, stressed/low stress/unstressed, slow/normal/fast, rhotic/non-rhotic, nasalized/non-nasalized, etc. It's a jungle in there and there are a lot of possible outcomes. How much prediction are you willing to pay for? – jlawler Oct 01 '16 at 16:34
1

I don't understand why neutralization of contrasts that exist in stressed vowels means schwa needs to be counted as a separate phoneme. Would you also say that speakers who merge intervocalic post-tonic /t/ and /d/ to [ɾ] have /ɾ/ as a phoneme? I perceive my pronunciations of "wetting" and "wedding" as the same: [ˈwɛɾıŋ]. If I just hear this form of the word, I can't predict which consonantal value would result if I removed the -ing suffix. – brass tacks Oct 01 '16 at 16:55
@sumelic, I don't think that schwa needs to be counted as a separate phoneme. All I said was that [ə] and stressed [ʌ] are (reflexes of) different phonemes, but I didn't mean to be saying what phonemes those are. The phonemes could be unstressed /ʌ/ (realized as [ə]) and stressed /ʌ/, for instance. – Greg Lee Oct 01 '16 at 17:17
2

I'm still having trouble understanding. I thought stress was suprasegmental, but you seem to be saying here it is a feature of the vowel phoneme. If unstressed and stressed /ʌ/ are different phonemes, and so on for any stressed-unstressed pair of vowels, doesn't that effectively double the phonemic inventory of English? – brass tacks Oct 01 '16 at 17:22
1

Yes, the premise of my answer was that stress is phonemic, which I take to mean that unstressed and stressed /ʌ/ are different phonemes, and that does increase the phonemic inventory, It's just like phonemic vowel nasality or vowel length increase the vowel phoneme inventories of those languages that have those phonemic distinctions. Or aspiration or labialization or palatalization do it for consonants. It's not unusual, – Greg Lee Oct 01 '16 at 17:32
1

"Stress is phonemic in English" But is it? I know it is not phonemic in Portuguese, where the stressed, unstressed, and reduced versions of a given vocalic phoneme are still just allophones. We know that because they morph in a regular way when words get suffixes. If stressed [ʌ/ə] is always /ʌ/ when stressed, and always /ə/ when unstressed, and if the change in stress caused by suffixation regularly makes /ʌ/ to morph into /ə/, then they are probably better considered as allophones of a phoneme. For instance, what happens to /ˈkʌs·təm/ when it becomes part of "customised"? – Luís Henrique Oct 02 '16 at 13:15
@LuísHenrique, Chomsky & Halle's The Sound Pattern of English is an extended argument that stress is not phonemic in English. So it's arguable. But I don't believe the analysis in SPE (though I think it's a lot of fun) -- for one thing, it relies on lots of non-phonetic underlying "phonemic" information. For instance, to predict that the noun "advocate" has no stress on the last syllable, unlike the verb "advocate", you have to rely on the "phonemic" part of speech, as well as morphological information about the "-ate" suffix. Also, the SPE analysis must recognize many exceptions. – Greg Lee Oct 02 '16 at 13:34
@LuísHenrique, Nothing happens to "custom" when "-ise/-ize" is added. The change of unstressed /ʌ/ to [ə] is regular in English, so if stress were predictable, perhaps the two sounds under discussion could considered allophones of one phoneme. However, stress in English is not predictable. – Greg Lee Oct 02 '16 at 13:46
1

@GregLee - Stress in not predictable in English - but the phonetic changes of vowels due to stress are (or so they seem to be in the case under discussion. Again with Portuguese: you have two different phonemes, /e/ and /ɛ/. /ɛ/ can only be stressed. But they are different phonemes because they form minimal pairs, /'fɛ.ʃu/ "I close" vs. /'fe.ʃu/ "lock". If the stressed vowel was always ɛ, I would say they were the same phoneme, with different allophones dependent on phonetic context. As, for instance, /a/, which is pronounced /ɐ/ when the consonant across the syllabic boundary is nasal, but... – Luís Henrique Oct 02 '16 at 14:22
... /a/ otherwise (leaving aside its pronounciation as a final vowel), never forming minimal pairs. – Luís Henrique Oct 02 '16 at 14:24
@LuísHenrique, I don't disagree with what you say. Phonologists do generally recognize that law-like relationships among phonemes are possible, as well as among allophones. Alternations among phonemes are sometimes called "morphophonemic", because it is common to find that they are conditioned by the class of morpheme involved. Your example of the relationship of open and close e in Portuguese is perhaps of this type. – Greg Lee Oct 02 '16 at 14:41
2

The strongest argument for schwa not being the same phoneme as /ʌ/ is still this comment by @chrylis: "schwa is in complementary distribution with all other vowels (except [ɪ])", which I take to mean that all vocalic phonemes, exception noted, turn into schwa when unstressed (which probably originates the widespread joke that English is spoken with a hot potato in mouth). If so, schwa cannot be the same phoneme as /ʌ/, but must be an allophone for every other English vowel. – Luís Henrique Oct 03 '16 at 13:49

score 1 · Answer 6 · answered Oct 08 '16 at 13:52

1

I don't know about American English but in standard English there is a minimal pair:

'Bacup' (a town in Lancashire) /beikəp/ and 'Bake up' /beik ʌp/, however the stress issue may still apply.

answered Oct 08 '16 at 13:52

Ned

606
4
8

Standard English? – Azor Ahai -him- Oct 08 '16 at 18:22
1

Yes - as spoken by the English in England – Ned Oct 10 '16 at 09:45
6

Then you should clarify that's what you mean. – Azor Ahai -him- Oct 10 '16 at 16:26
English “as spoken by the English is England” is not “standard English” – it’s British English, or more specifically English English (though it sounds daft). And it’s certainly not just one English. – Janus Bahs Jacquet Apr 30 '20 at 05:33

score 1 · Answer 7 · answered Dec 20 '18 at 21:07

This is just my observation but it's in favor of this hypothesis. In songs for example where the meter clashes with the natural stress of a word containing a schwa, I have observed that the stressed schwa becomes /ʌ/, even when the spelling of the specific word does not indicate that there ever was a /ʌ/ in this position. (Maybe I'll be able to recall an example later.)

In addition to that I would call the articulation of /ʌ/ the most similar to /ə/ from all English phonemes.

score 0 · Answer 8 · answered Apr 10 '19 at 20:01

To me, the schwa is similar to the Turkish dotless i. A vowel close to no real differentiation, and with that imo a shorter and less pronounced version of ir as in birth. Once a vowel is pronounced in a very 'flat' way and it gets short (as when there is less stress on it), we get the schwa - with the above mentioned exception /ɪ/.

The word custom is actually a good way to show the difference. If we replace the schwa with an /ʌ/, we get something like custum (let's put stress on both syllables for clarity, but with a little bit of trying, one should be able to just stress the first syllable and still get the u-sound, not the schwa, in the unstressed syllable).

The u in backup is also a good example. If we pronounce it the same as back up (trying to stress both words, for clarity again), we should hear a difference to the American way of saying backup (trying to stress the up but keep the pronunciation 'flat').

In both words, we can try to stretch out the schwa or the /ʌ/, and we will hear a sound similar to ir in birth and a in car, respectively.

American English tends to 'flatten' the different vowels, so a lot of vowels could probably be replaced with the schwa. I'd leave it in the current way in IPA, because a lot of people do actually have a hint of the vowel in the pronunciation, with the phoneme chosen according to the usual rules or regional variation.

Where it's rare to pronounce the vowel clearly, one might also choose the second most popular pronunciation in the region or in important dialects (British and American English, in the future probably also Indian English), with some indicator that people often 'flatten' it to a schwa.

One might also consider that pronunciation is often not only affected by the spelling of a word, but also the way dictionaries say they are supposed to pronounce it. Putting a schwa everywhere would probably not help people pronounce things so that an English speaker from another country can understand them. While having a 'real' vowel there would give people a 'standard pronunciation' for when trying to talk especially clear, for instance to a foreigner.

The thing with the dictionary is… While many a dictionary differentiate between schwa or chevron, others do not. Some that consider them a merger: The Kindle's dictionary (The New Oxford American Dictionary, e.g.: “someone” = /ˈsəmˌwən/); dictionary.com (e.g.: suhm-wuhn); and aforementioned CMU, used extensively in applied linguistics. There seems to be an American preference in favor of the merger and a British tradition ("a schwa is never stressed!") in favor of a split… though my question is specifically about American English. — Fabien Snauwaert, Apr 10 '19 at 22:07

American English : are [ə] and [ʌ] different phonemes? (schwa vs. chevron)

8 Answers8