Beyond Phoneme Inventories: Building a complete phonology

[Sumelic]

OK, so this is going to be a thread about all kinds of phonology and morphophonology topics, with a focus on other things besides phoneme inventories. Some topics I might post about are phonotactics, allophony, neutralizations, metathesis, suprasegmentals and long distance processes. I probably won't get into prosody, unless I learn a lot more about it and find something interesting to say.

I won't really be making any kind of order to these, at least not at this point, I'll just be posting about things I think are interesting or maybe inspirational from natural languages.

Also, if anybody else wants to post something cool here, that would be great !

[Sumelic]

For my first topic, I'll be talking about allophony (Did you know it's pronounced /əˈlɑfəni/? I didn't until today) between voiced lateral and central liquids like [l], [r] and [ɾ], and sometimes even the plosive [d], in different languages. A number of languages have a single liquid phoneme that can be realized as either central or lateral; sometimes this is conditioned by the phonetic environment.

We'll start out with some examples from the Bantu languages.

In Luganda, /l/ usually has the allophone /ɾ/ after /e/, /eː/, /i/ or /iː/. This could be analysed as /l/->[ɾ]/_[+front] or /l/->[ɾ]/_[+close][-rounded]. However, these two allophones can also substitute for one another freely in all contexts. When prenasalized (preceded by a homorganic syllabic nasal), it merges with the usually distinct phoneme /d/. When geminated, it is realized as /dː/, the same as the geminate of /d/.

Sesotho has a phonemic plosive inventory of /pʼ pʰ b tʼ tʰ kʼ kʰ/.
However, /l/ is realized as [d] before a high vowel /i/ or /u/. This is believed to have developed from [ɽ]. This pattern of allophony also exists in the related language Tswana, which also has a distinct alveolar trill phoneme /r/. Sesotho also had /r/ historically, but the modern realization of this phoneme tends to be /ʀ/, possibly due to French influence.

Many Polynesian languages have only one liquid. In Samoan, /l/ is generally pronounced as a central flap [ɾ] when it is both following a back vowel /a, o, u/ and preceding the high front vowel /i/.

The language that most famously has l~r allophony is probably Japanese. Japanese /r/ may be realized in a variety of ways: as a lateral approximant, a lateral or central flap, or sometimes even an alveolar trill. It tends to be realized as a central flap [ɾ] most often before /i/ and /j/, and it is most likely to be realized as a lateral approximant /l/ before /o/. In general, though, the two sounds are in free variation with each other and other realizations of the phoneme.

In the nearby language Korean, /l/ is realized intervocalically as an alveolar flap [ɾ], and as the lateral approximant [l] or [ɭ] at the end of a syllable or when following another /l/. In South Korea, this sound cannot occur at the start of a word or directly following another consonant: historical /l/ in these positions generally merged with /n/ (followed by later simplification of /nj/ and /ni/ to /j/ and /i/). However, in the sequence /nl/, the n is assimilated with the result of a geminate /ll/. Loanwords from English have now reintroduced the phoneme word-initially, where it exhibits free variation between [ɾ] and [l].

Hopefully some of these examples will appeal to you when you are making your language's phonology, or inspire you to create your own pattern of allophony involving these sounds.

[GrandPiano]

(Did you know it's pronounced /əˈlɑfəni/? I didn't until today)

What?! And I've been pronouncing it /ˈæləfɵni/ this whole time!

/əˈlɑfəni/ just feels so "wrong"… Like, it doesn't even feel like the same word…

[Sumelic]

Well, it seems to follow the pattern of "cacophony" and "polyphony", and other words suffixed with -phony. The "o" was actually long in Greek, and presumably would have also been long in Latin, but I think we got this family of words through French. Actually Wiktionary does give /ˈæləfoʊni/ as an (unsourced) alternative pronunciation, which is actually how I used to say it before today, but to me that looks suspiciously like a back-formation from "allophone", and I haven't been able to find that pronunciation anywhere else except a Youtube "How do you pronounce X" video that sounded rather robotic. It seems safer to go with the antepenult. It looks like this is also the pronunciation used by John C. Wells, although he seems to suggest he has heard other pronunciations of the word.

[Znex]

I think it's also the Brit Commonwealth way to pronounce four-syllable words on the antepenult than on the penult/first syllable. So I don't think there's really anything wrong with pronouncing allophony either way; in a way, it's a sort of allophony itself.

(But it's not, because it's free distribution.)

[Sumelic]

Sure, people can say it however they want. I don't think that's a general rule about British/American pronunciation differences, but nobody's going to jump on you if you stress a different syllable.
P.S. Free distribution is also a kind of allophony

[DesEsseintes]

In South Korea, this sound also cannot occur in the syllable onset; historical /l/ in this postion merged with /n/

Perhaps you mean word-initially? /l/ is definitely found in syllable onsets word-medially, as evidenced by the grammatical particles -를 -leul and -로 -lo, and pronounced according to the allophonic principles you laid out.

[Sumelic]

Yeah, I messed it up. I don't know any Korean . Word-initially is how I found the rule described, but it seems like /l/ also merged with /n/ within a word after any other obstruent, and I was trying to include that information also. If after a vowel, it would be realized as [r], right? So I guess what I was trying to say was that [l] would never occur in the syllable onset in native South Korean words, but I messed up and said /l/ instead. Also, I forgot about when it comes after another /l/, in which case it would just be geminate /ll/, the first in the coda of one syllable and the second in the onset of the next. Does that sound right now?

[Sumelic]

Actually, could I also ask about one other detail, DesEsseintes? What happens in t+l clusters and n+r clusters? It looks like nr > ll is agreed on by all of my sources. Wikipedia has a chart indicating tl > nn, but I found a paper giving the rule tl > ll.

[Xing]

For my first topic, I'll be talking about allophony (Did you know it's pronounced /əˈlɑfəni/? I didn't until today) between voiced lateral and central liquids like [l], [r] and [ɾ], and sometimes even the plosive [d], in different languages. A number of languages have a single liquid phoneme can be realized as either central or lateral; sometimes this is conditioned by the phonetic environment.

Before talking about specific examples, I think it might be a good idea to define terms like 'phone', 'phoneme', 'allophone', 'complementary distribution', 'free variation', etc.

The next thing I would do would probably be to describe the kinds of phonological processes that can be at work in allophony - such as fortition, lenition, various kinds of assimilation, etc.

Btw, I think I'd say /əlˈɒfəni/ - in analogy with similar Greek-derived words...

[Sumelic]

Allright. I'm working on an overall analysis of these terms, but for now, I'll make posts with individual definitions and see if people think I've missed anything. Once I've developed good definitions, I'll edit them into the first post.

Phones: the sounds that come out of somebody’s mouth when they speak. The only way to know what these are precisely is to hear somebody speak, or to have a machine record the sound for us and then analyze it with a program like Praat. For convenience, we sometimes use the International Phonetic Alphabet (IPA) to give simplified written descriptions of utterances: when IPA is used like this, it is conventional to enclose the phones in [square brackets]. The way the IPA works is that sets of phones that sound very similar to one another are grouped together and represented by a single symbol/letter. The IPA also makes use of diacritics on letters and sequences of letters to transcribe phones. In general, the full IPA with letters, diacritics and digraphs is adequate for representing all of the contrasting sounds that can be found in natural languages. An IPA transcription can give you the general idea of what something sounds like, but even at its most detailed, it can’t give you all the specifics, and people frequently leave out details that are predictable, difficult to type, or that use rarer symbols or that don't have symbols. When a minimal amount of detail is included, it is called “broad transcription”. When more detail is included (say, if a word is transcribed with enough detail for someone who doesn’t know the language but knows IPA to be able to pronounce it in a clear manner, if with an accent) it is called a “narrow transcription”. (If you don't know IPA and would like to learn, you might find the following IPA and phonetics resources helpful.) My transcriptions in this thread will vary in how broad they are.

[Serena]

In South Korea, this sound cannot occur at the start of a word or directly following another consonant: historical /l/ in these positions generally merged with /n/

Well, not quite.

1) First of all, it does not merge with /n/. It is indeed realized as /n/, but in a position whereas the <n> is not pronounced /n/ at all. Summing up:

• Word-initial <n ㄴ> is an unvoiced nasal alveolar stop.
• Word-initial <r ㄹ> is a voiced nasal alveolar stop /n/.

This is indeed not a case of merging.

2) This phenomen only applies to words of chinese origin, and most pure korean words won't begin with <r>. For example, the word for "woman" nǘzǐ has been migrated to ryeoja 려자 to imitate that sound.

This chinese n becomes a r because initial consonants in korean are never voiced (not even the nasals), except for the r, and using that letter was the best device they thought of to convey that voicedness. Eventually, the r slowly decayed and "woman" because simply yeoja 여자.

[GrandPiano]

For example, the word for "woman" nǘzǐ has been migrated to ryeoja 려자 to imitate that sound.

女子 is nǚzǐ, not nǘzǐ. (It's pronounced nǘzǐ because of tone sandhi, but the underlying form is nǚzǐ)

[clawgrip]

Another somewhat interesting type of allophony related to this topic is that for many speakers, Korean initial nasals (/n/ and /m/) are denasalized. I'm not particularly knowledgeable about the topic, but I think it's a relatively recent phenomenon. I first came across this when I noticed that several Koreans I interacted with on a mostly daily basis sounded like they had colds because they never pronounced /n/ and /m/ properly (in English). I decided to look it up, and sure enough, there are several things on the Internet about it.