Page 1 of 2

What is the limit of similarity for allophones?

Posted: Sun Jul 11, 2010 12:05 pm
by Kai_DaiGoji
Or let me try a different way to ask this question - how different can two sounds be and yet still have some language consider them allophones? I know what allophones are, and that different languages will see certain sounds as the same, etc. But for example, the alveolar tap [ɾ] is considered an allophone of [t] or [d] in English (I think, like in 'better' right?) but it's considered, well, not an allophone but some kind of 'r' sound in spanish. [t] is clearly nothing like [r] but there you go. Or in English the glottal stop is often considered an allophone of [t] (or in my case, [nt]).

So I guess what I'm asking, is it attested or possible for sounds that are not similar at all to still be considered allophones? For not just [p] and but also [m] to be allophones ? Or all the nasals? Or [k] and [v]?

Posted: Sun Jul 11, 2010 12:13 pm
by Soap
Well, in a language where PoA isn't distinctive in syllable codas, all nasals could be allophones of each other in syllable final position. Japanese does this, and adds nasalized /j/ and /w/ into the mix as well. It could also be argued that the Japanese syllable-final nasals aren't actually allophones, they're just the normal phonemes /m/, /n/, etc occurring in restricted environments. But that would still leave us with similar processes in Spanish and many other languages.

Posted: Sun Jul 11, 2010 12:26 pm
by Mbwa
Since allophones tend to come from historical processes, I'd say that almost any two phones could be allophones of each other.

Posted: Sun Jul 11, 2010 12:37 pm
by Morrígan
I'd argue that [t], [r], and [ɾ] are pretty dang similar:

Code: Select all

             t  r  ɾ

Sonorant     0  1  0
Continuant   0  1  0
Distributed  1  0  1
Voice        0  1  1
I guess the interesting question might be at what point phonetic dissimilarity overrides the clues of statistical patterning.

Posted: Sun Jul 11, 2010 12:51 pm
by Aurora Rossa
Since allophones tend to come from historical processes, I'd say that almost any two phones could be allophones of each other.
I dunno if you can stretch allophony that far. We don't consider [h] and [N] allophones in English even though they have complementary distribution. I've always heard the reason is because they sound too different from each other.

Posted: Sun Jul 11, 2010 1:05 pm
by Mbwa
Eddy wrote:
Since allophones tend to come from historical processes, I'd say that almost any two phones could be allophones of each other.
I dunno if you can stretch allophony that far. We don't consider [h] and [N] allophones in English even though they have complementary distribution. I've always heard the reason is because they sound too different from each other.
What I meant was that, in the /t/~[4] example, [t] weakened to [4] between vowels. So, even if you don't consider the phones similar, the [4] came from a sound change to the [t]. There wasn't a change [h] > [N] / _#.

So like, a language could have [v] as an allophone of /k/, maybe due to something like:

k > G / V_V
G > v / _u

Posted: Sun Jul 11, 2010 1:06 pm
by Morrígan
Eddy wrote:
Since allophones tend to come from historical processes, I'd say that almost any two phones could be allophones of each other.
I dunno if you can stretch allophony that far. We don't consider [h] and [N] allophones in English even though they have complementary distribution. I've always heard the reason is because they sound too different from each other.
They also don't experience morphological alternation, which is the other really big (possibly the bigger) clue regarding underlying forms.

Posted: Sun Jul 11, 2010 4:49 pm
by Magb
Hawai'ian famously has [t] as an an allophone of /k/, though as I understand it it may be more of a case of free variation than complementary distribution.

Many Norwegian and Swedish dialects have [r`] as an allophone of /l/ (though at least in my dialect /r`/ is actually a marginal phoneme). [l] and [r`] aren't exactly birds of a feather.

Posted: Sun Jul 11, 2010 5:02 pm
by finlay
Magb wrote:Hawai'ian famously has [t] as an an allophone of /k/, though as I understand it it may be more of a case of free variation than complementary distribution.

Many Norwegian and Swedish dialects have [r`] as an allophone of /l/ (though at least in my dialect /r`/ is actually a marginal phoneme). [l] and [r`] aren't exactly birds of a feather.
yeah they are.

Posted: Sun Jul 11, 2010 5:07 pm
by Magb
finlay wrote:yeah they are.
I assume you're referring to [l] and [r`]? I disagree, but I'd be happy to be proven wrong with some spectrograms.

Posted: Sun Jul 11, 2010 5:32 pm
by finlay
they are birds of a feather because they are treated in a similar way cross-linguistically ... r and l are famously interchangeable. thus it does not surprise me that a 'rhotic' might be in a 'lateral' phoneme. but i guess you are right in that the rhotic/liquid group arguably does not have much phonetic similarity between members and it is very much a phonological grouping

but i'm not surprised, anyway. at the same time, the "phonetic similarity" rule always struck me as the most arbitrary when defining phonemes. things which are deemed too unsimilar in another language might be similar enough to be the same phoneme in another language

Posted: Sun Jul 11, 2010 5:47 pm
by Magb
I guess you can approach the question in the thread title in two different ways: allophones that are phonetically dissimilar (whatever that may mean), or allophones that share few phonological features. For instance the [t~k] alternation in Hawai'ian is striking because /t/ and /k/ tend to be quite phonetically distinct in languages that have both -- or at least viewed as being distinct by speakers of said languages -- but phonologically speaking they only differ in their point of articulation.

I think you're right that phonetic differences aren't necessarily important, but phonetically dissimilar allophones will tend to be surprising to speakers of languages where the allophones are not only viewed as distinct phonemes, but as strongly distinct phonemes. As you say, "similarity" is in the ears of the beholder.

Posted: Sun Jul 11, 2010 7:11 pm
by vohpenonomae
Eddy wrote:
Since allophones tend to come from historical processes, I'd say that almost any two phones could be allophones of each other.
I dunno if you can stretch allophony that far.
Yeah, you can; in one or another Algonquian language, through merger and assimilation, these allophonic variants hold:

č:n

š:n

ł:y

ł:n

Posted: Sun Jul 11, 2010 7:35 pm
by Radius Solis
t~k is only the tip of the iceberg in Hawaiian. In that language ALL of the following phones are possible as realizations of "/k/": [t T s ts S tS d D z dz Z dZ k x g G]. Plus many in-between sounds. The trick is to understand that this phoneme is specified for only two features: [+lingual +obstruent]. Any non-sonorant sound you can make with the tongue counts as /k/ in Hawaiian, it doesn't matter if it's voiced or what its exact POA or MOA is. Retroflexes, uvulars, whatever, it's all /k/.

Posted: Sun Jul 11, 2010 8:01 pm
by zompist
Historically, very distinct allophones generally derive from 'the same sound', or more precisely, from more similar sounds. So there's no theoretical limit on how distinct they can be; they're only constrained by there being a path from the original sound, and there are few if any restraints on that.

English h/ng are by no means impossible as allophones; we just have no evidence that they have a common origin, and native speaker intuition certainly considers them separate.

Posted: Sun Jul 11, 2010 10:22 pm
by Torco
zompist wrote:English h/ng are by no means impossible as allophones; we just have no evidence that they have a common origin, and native speaker intuition certainly considers them separate.
Which kind of is the point of allophones, right? stuff that, to the speakers, sounds kind of the same ?

Posted: Sun Jul 11, 2010 11:00 pm
by Kai_DaiGoji
Torco wrote:
zompist wrote:English h/ng are by no means impossible as allophones; we just have no evidence that they have a common origin, and native speaker intuition certainly considers them separate.
Which kind of is the point of allophones, right? stuff that, to the speakers, sounds kind of the same ?
But, if I'm following correctly, even though they aren't in english, there's no reason h/ng couldn't be allophones in some languages, right?

Posted: Sun Jul 11, 2010 11:38 pm
by Morrígan
Kai_DaiGoji wrote:
Torco wrote:
zompist wrote:English h/ng are by no means impossible as allophones; we just have no evidence that they have a common origin, and native speaker intuition certainly considers them separate.
Which kind of is the point of allophones, right? stuff that, to the speakers, sounds kind of the same ?
But, if I'm following correctly, even though they aren't in english, there's no reason h/ng couldn't be allophones in some languages, right?
Right. But they are also more likely to be allophones in some environments rather than others. Which it may or may not share with other nasals.

{n m ŋ } > h _C[–Voice]

The above apparently is rather common in languages of North America.

Posted: Mon Jul 12, 2010 12:20 am
by schwhatever
Another natural example: aspirate voiceless plosives ([t_h] for instance). To the average English-speaker the bold parts of top and pretty are analyzed as being the same (ie, we can't recognize the difference between [t_h] and [t]). To a Hindi-speaker, however, they're clearly separate phonemes /t t_h/.

It's all about context.

Posted: Mon Jul 12, 2010 2:43 am
by Niedokonany
IME what speakers think of such sounds in complementary distribution is greatly influenced by the writing system they use. You can run into complicated cases like Polish i/y, which 1) are largely in complementary distribution, changing one into another depending on the preceding consonant 2) are distinguished by native speakers, as they are distinguished by the official orthography, and generally can be said in isolation 3) rhyme in poetry with one another 4) a few near-minimal pairs do appear because of loanwords. Personally, I prefer to consider them to be separate phonemes, but 3) is kind of odd, isn't it?

Posted: Mon Jul 12, 2010 6:28 am
by AnTeallach
In Scouse I understand that some words (like that) can be pronounced with the last consonant either [h] (utterance finally) or [4] (before a word beginning with a vowel). This looks a lot like [4] and [h] being allophones of the same phoneme.

(Paper here, though the encoding of some IPA symbols is messed up.)

Posted: Mon Jul 12, 2010 7:40 am
by Yng
AnTeallach wrote:In Scouse I understand that some words (like that) can be pronounced with the last consonant either [h] (utterance finally) or [4] (before a word beginning with a vowel). This looks a lot like [4] and [h] being allophones of the same phoneme.
That sounds about right. Scouse is the only dialect of English that has [h] finally that I can think of.

Re: What is the limit of similarity for allophones?

Posted: Mon Jul 12, 2010 7:08 pm
by TomHChappell
Kai_DaiGoji wrote:Or let me try a different way to ask this question - how different can two sounds be and yet still have some language consider them allophones? I know what allophones are, and that different languages will see certain sounds as the same, etc. But for example, the alveolar tap [ɾ] is considered an allophone of [t] or [d] in English (I think, like in 'better' right?) but it's considered, well, not an allophone but some kind of 'r' sound in spanish. [t] is clearly nothing like [r] but there you go. Or in English the glottal stop is often considered an allophone of [t] (or in my case, [nt]).

So I guess what I'm asking, is it attested or possible for sounds that are not similar at all to still be considered allophones? For not just [p] and but also [m] to be allophones ? Or all the nasals? Or [k] and [v]?
Aren't there natlangs in which all close (high) vowel-phones are allophones of the same "close vowel" phoneme, and likewise all open (low) vowel-phones are allophones of the same "open vowel" phoneme?

Re: What is the limit of similarity for allophones?

Posted: Mon Jul 12, 2010 7:20 pm
by Morrígan
TomHChappell wrote:
Kai_DaiGoji wrote:Or let me try a different way to ask this question - how different can two sounds be and yet still have some language consider them allophones? I know what allophones are, and that different languages will see certain sounds as the same, etc. But for example, the alveolar tap [ɾ] is considered an allophone of [t] or [d] in English (I think, like in 'better' right?) but it's considered, well, not an allophone but some kind of 'r' sound in spanish. [t] is clearly nothing like [r] but there you go. Or in English the glottal stop is often considered an allophone of [t] (or in my case, [nt]).

So I guess what I'm asking, is it attested or possible for sounds that are not similar at all to still be considered allophones? For not just [p] and but also [m] to be allophones ? Or all the nasals? Or [k] and [v]?
Aren't there natlangs in which all close (high) vowel-phones are allophones of the same "close vowel" phoneme, and likewise all open (low) vowel-phones are allophones of the same "open vowel" phoneme?


Well, I think Kabardian is only supposed to have low /a/ and nonlow /ə/. Fairly common in West Caucasian. One of them (possibly also Kabardian) might have long /aː/. But, still it's low-nonlow or low-high distinction.

Re: What is the limit of similarity for allophones?

Posted: Mon Jul 12, 2010 7:30 pm
by TomHChappell
TheGoatMan wrote:Well, I think Kabardian is only supposed to have low /a/ and nonlow /ə/. Fairly common in West Caucasian. One of them (possibly also Kabardian) might have long /aː/. But, still it's low-nonlow or low-high distinction.
Thanks for the example.

Aren't there natural languages in which all (or nearly all) voiceless fricative phones are allophones of the same phoneme? So that, for instance, maybe [f h s S T W] are all allophones of what might as well be written /h/?