English coda 'rhinoglottophilia' revisited

[Zju]

Honest question: are there any reasons other than historical and phonetical to not analyse [h] and [ŋ] as allophones of the same phoneme? They're in complementary distribution and all.
For the sake of the thought experiment, let's say OE [nɡ] developed differently thus: nɡ > ŋɡ > ŋɣ > ɣ, would then [h] and [ɣ] be considered the same phoneme, despite being in complementary distribution by chance and their non trivial would-be allophony? What if the sound change went a further step to [x] or [ɦ]?

[jal]

Honest question: are there any reasons other than historical and phonetical to not analyse [h] and [ŋ] as allophones of the same phoneme?

\
I would think that diachronics and phonetics are about the only tools one has to analyse a language's phonology (esp. if including knowledge on actual sound changes and allophones in the world's languages with "phonetics")? What other reason could there be? Even when talking about complementary distribution, you are talking phonetics.


JAL

[sangi39]

They're in complementary distribution and all.

Are they, though? At least in my speech both can appear between vowels.

[Travis B.]

They're in complementary distribution and all.

Are they, though? At least in my speech both can appear between vowels.

Same here (e.g. [ˈsẽːŋʁ̩(ː)] versus [ˈkʰʌːɾɘhe(ː)]). To discount this for me, one would have to say that names don't count for phonemic analysis, which is a stretch.

To me the two differ because /ŋ/ conditions a long preceding vowel while /h/ conditions a short preceding vowel (including across word boundaries), and trying to unify the two would result in a phoneme that conditions both depending on position, unlike any other phoneme.

[Richard W]

There doesn't seem to be any reason not to treat them as different. One can also find idiolects where nonsense words force them to be separate. for example, there are native Enɡlish speakers for whom <singer> is /sɪ.ŋər/ and to whom word-initial /ŋ/ comes easily. Another is that when trying to add a word-final /h/ to their repertoire, speakers do not normally fail by appending [ŋ] instead.

There are languages where the assignment of phones to phonemes is arbitrary. The assignment of the [p] of English <spin> to /p/ rather than /b/ is generally taken to be arbitrary, though it may actually have a basis. For most English speakers, literacy precedes (generally pre-puberty) (near-peak) mastery of the grammar (early teens?), and to ignore the effects of literacy is dogmatic. However, one should be very wary of claiming that /sb/ does not occur at the start of English words.

I believe you will find that most native speakers of English who understand the concepts find the idea of [ŋ] and [h] being allophones ridiculous. Minimal pairs demonstrate that two phones belong to different phonemes, but, for example, near-minimal pairs should not be ignored.

[Porphyrogenitos]

The introductory linguistics textbook I was reading today specifically states that while the existence of minimal pairs does indicate that two phones are phonemic, the lack of a minimal pair does not mean that they are not phonemic. It's just a convenient shortcut for figuring out if two sounds are phonemic. Factors such as major phonetic difference must also be considered.

[Chengjiang]

Honest question: are there any reasons other than historical and phonetical to not analyse [h] and [ŋ] as allophones of the same phoneme? They're in complementary distribution and all.

To give another reason why not: There is no morphological or phonological process connecting them. If there were a process that converted morphemes with [h] to morphemes with [ŋ] or vice versa in contemporary English, that would be one thing, but there isn't.

Also, as others have pointed out both segments can occur intervocalically in most (not all!) dialects of English. In order to analyze them as being in fully complementary distribution one would have to treat all intervocalic [h] as underlyingly an onset and all intervocalic [ŋ] as underlyingly a coda, and I honestly can't think of a way to do this for most dialects without doing the very thing you want to avoid, i.e. bringing in diachronics.

The introductory linguistics textbook I was reading today specifically states that while the existence of minimal pairs does indicate that two phones are phonemic, the lack of a minimal pair does not mean that they are not phonemic. It's just a convenient shortcut for figuring out if two sounds are phonemic. Factors such as major phonetic difference must also be considered.

This is a very important distinction. If p implies q, it does not then follow that NOT p implies NOT q.

Also, more subtly, the absence of minimal pairs does not automatically mean complementary distribution. Two phones can occur in the same or overlapping environments while a pair of otherwise identical words that differ in which of the two phones they have may happen not to exist in the lexicon.

[Sumelic]

There doesn't seem to be any reason not to treat them as different. One can also find idiolects where nonsense words force them to be separate. for example, there are native Enɡlish speakers for whom <singer> is /sɪ.ŋər/ and to whom word-initial /ŋ/ comes easily.

Really? That's surprising. I've never heard of it.

Another is that when trying to add a word-final /h/ to their repertoire, speakers do not normally fail by appending [ŋ] instead.

That's a good point.

There are languages where the assignment of phones to phonemes is arbitrary. The assignment of the [p] of English <spin> to /p/ rather than /b/ is generally taken to be arbitrary, though it may actually have a basis. For most English speakers, literacy precedes (generally pre-puberty) (near-peak) mastery of the grammar (early teens?), and to ignore the effects of literacy is dogmatic. However, one should be very wary of claiming that /sb/ does not occur at the start of English words.

I've definitely thought that this is arbitrary. I think there is some evidence for it being /sb/, such as the analogous devoicing that occurs for some speakers in words like "disgust," and for all speakers in the past tense suffix. But I also found some evidence for it being phonemically /sp/, such as the fact that speakers of some dialects, and some children that are acquiring English, drop the [s] and realize it as aspirated [pʰ] (merging the cluster with /p/). So as you say, /sp/ may also have a basis.

Also, as others have pointed out both segments can occur intervocalically in most (not all!) dialects of English. In order to analyze them as being in fully complementary distribution one would have to treat all intervocalic [h] as underlyingly an onset and all intervocalic [ŋ] as underlyingly a coda, and I honestly can't think of a way to do this for most dialects without doing the very thing you want to avoid, i.e. bringing in diachronics.

Syllabifying them that way doesn't seem at all unreasonable to me, even synchronically. It's the distribution found in monosyllabic words, and also [ŋ] traditionally can only be preceded by a lax vowel (although this is no longer true in some American accents), which implies that it is the coda of the preceding syllable.

[Travis B.]

Also, as others have pointed out both segments can occur intervocalically in most (not all!) dialects of English. In order to analyze them as being in fully complementary distribution one would have to treat all intervocalic [h] as underlyingly an onset and all intervocalic [ŋ] as underlyingly a coda, and I honestly can't think of a way to do this for most dialects without doing the very thing you want to avoid, i.e. bringing in diachronics.

Syllabifying them that way doesn't seem at all unreasonable to me, even synchronically. It's the distribution found in monosyllabic words, and also [ŋ] traditionally can only be preceded by a lax vowel (although this is no longer true in some American accents), which implies that it is the coda of the preceding syllable.

I personally would avoid phonemic syllabification if at all possible, and in this case, prefer an analysis where there's two separate phonemes over one where there's one phoneme which varies depending on phonemic syllabification.

[Sumelic]

I personally would avoid phonemic syllabification if at all possible, and in this case, prefer an analysis where there's two separate phonemes over one where there's one phoneme which varies depending on phonemic syllabification.

I wouldn't analyze them as one phoneme, but that doesn't have to follow from them being in complementary distribution. For example, isn't /h/ in complementary distribution with any of the lax vowels? I can't think of a minimal pair for /h/ and /ɪ/. Are you generally opposed to analysis where contrasts are derived from syllabification (rather than segmental differences), or do you just think there's not enough evidence in this particular case?

To give another reason why not: There is no morphological or phonological process connecting them. If there were a process that converted morphemes with [h] to morphemes with [ŋ] or vice versa in contemporary English, that would be one thing, but there isn't.

I feel like morphophonological evidence is the strongest indicator of allophony, but it doesn't always exist. For example, my speech exhibits raising of TRAP (which I'll transcribe as [æ] in general) to something like [eə] before nasals, entirely regularly. I can't think of any pair of related words where one has [æ] and the other has [eə], since almost no morphemes end in TRAP. But despite the lack of morphological evidence, it doesn't make sense to analyze my [eə] as a distinct phoneme, since it only occurs in restricted environments and is always in complementary distribution with another phoneme or set of phonemes (/æ/ before nasals; /eɪ/ before tautosyllabic /l/; /æ/, /ɛ/ and /eɪ/ before tautosyllabic /r/).

One case that I think might somewhat reasonably be analyzed as allophony, but where the phones come from historically distinct sources, is the syllable-initial consonant cluster [ʃr]. In my native vocabulary, [ʃr] and [sr] do not contrast, so it is possible to analyze [ʃr] as /sr/, with [ʃ] as an allophone of /s/ before /r/. This has some historical support, like the word "grocery," but for the most part /ʃr/ comes from historical Germanic /skr/, and in other cases /sk/ became the phoneme /ʃ/, so the overall history would suggest an analysis as /ʃr/. But there are other synchronic processes that support [ʃr] = /sr/; it seems analogous to the broadly accepted identification of [tʃr] and [dʒr] (for speakers who have affricates like these in words like "train" and "dream") with phonemic /tr/ and /dr/, and some speakers actually use something closer to phonetic [sr] in words like "shrimp."

[Travis B.]

I personally would avoid phonemic syllabification if at all possible, and in this case, prefer an analysis where there's two separate phonemes over one where there's one phoneme which varies depending on phonemic syllabification.

I wouldn't analyze them as one phoneme, but that doesn't have to follow from them being in complementary distribution. For example, isn't /h/ in complementary distribution with any of the lax vowels? I can't think of a minimal pair for /h/ and /ɪ/. Are you generally opposed to analysis where contrasts are derived from syllabification (rather than segmental differences), or do you just think there's not enough evidence in this particular case?

The problem with introducing phonemic syllabification in the case of English is that otherwise there is little to no evidence of it, and one cannot just introduce phonemic syllabification in one place without making syllabification everywhere be phonemic. Also, yes, I would oppose the introduction of phonemic syllabification where the same contrast can be explained through distinct segments and there is not overwhelming evidence for phonemic syllabification.

Trying to explain this contrast in terms of syllabification or in terms of laxness versus tenseness of preceding vowels raises the question which has been raised before: just how does this play into a purported phonological process that results in [h] in some cases and [ŋ] in others? In this case, the answer is that there is no evidence of such a phonological process.

To give another reason why not: There is no morphological or phonological process connecting them. If there were a process that converted morphemes with [h] to morphemes with [ŋ] or vice versa in contemporary English, that would be one thing, but there isn't.

I feel like morphophonological evidence is the strongest indicator of allophony, but it doesn't always exist. For example, my speech exhibits raising of TRAP (which I'll transcribe as [æ] in general) to something like [eə] before nasals, entirely regularly. I can't think of any pair of related words where one has [æ] and the other has [eə], since almost no morphemes end in TRAP. But despite the lack of morphological evidence, it doesn't make sense to analyze my [eə] as a distinct phoneme, since it only occurs in restricted environments and is always in complementary distribution with another phoneme or set of phonemes (/æ/ before nasals; /eɪ/ before tautosyllabic /l/; /æ/, /ɛ/ and /eɪ/ before tautosyllabic /r/).

The difference here is that the two can be easily phonologically linked with one another, unlike [h] and [ŋ].

One case that I think might somewhat reasonably be analyzed as allophony, but where the phones come from historically distinct sources, is the syllable-initial consonant cluster [ʃr]. In my native vocabulary, [ʃr] and [sr] do not contrast, so it is possible to analyze [ʃr] as /sr/, with [ʃ] as an allophone of /s/ before /r/. This has some historical support, like the word "grocery," but for the most part /ʃr/ comes from historical Germanic /skr/, and in other cases /sk/ became the phoneme /ʃ/, so the overall history would suggest an analysis as /ʃr/. But there are other synchronic processes that support [ʃr] = /sr/; it seems analogous to the broadly accepted identification of [tʃr] and [dʒr] (for speakers who have affricates like these in words like "train" and "dream") with phonemic /tr/ and /dr/, and some speakers actually use something closer to phonetic [sr] in words like "shrimp."

The reason why [tʃr] and [dʒr] are treated as /tr/ and /dr/ is that there still are speakers with unaffricated pronunciations, or affricated ones that differ from a mere sequence of /tʃ/ or /dʒ/ and /r/.