Pitch Accent

[holbuzvala]

So I have a language I'm working on (surprise, surprise), and as a design goal I'm interested to have a pitch-accent system kind of like Japanese in it.

General Qs:
1. What other languages have comparable systems?
2. Are there particular resources you'd recommend for learning about such?

Spec Qs:
3. What are the effects of certain sounds on pitch? For instance, would velars/uvulars raise/lower pitch? What about the effect, instead of POA, of ejectivity, voicing, etc.?

All pointers welcome.

[Pabappa]

I dont think the consonants would interact with the pitch of the syllables. I dont really have anything else to add but just wanted to get rid of this idea up front .... of all the natlangs we have sound changes for, Im not aware of even one in which tone interacts with consonants.

[Vijay]

Um what? It usually does interact with consonants AFAIK. See Middle Chinese to any of the Chinese languages...

In Thai, for example, consonants even affect tone synchronically, and no syllable with an obstruent in coda position can have mid tone.

[Kuchigakatai]

1. What other languages have comparable systems?

Naturally, the most similar system to that of Tokyo/Standard Japanese would be that of Kyoto Japanese. Its phonology combining phonemic tone and phonemic pitch accent is pretty interesting: every content word must have a phonemic tone on the first syllable, and a phonemic accent may or may not be found on a subsequent syllable. Classic examples: /a˥me/ [a˥me˥] 'rain' (high tone, no accent), /ha˥ꜜna/ [ha˥na˨] 'flower' (high tone, accent present), /na˨deꜜɕiko/ [na˨de˥ɕ(i˨)ko˨] (low tone, accent on second syllable) '(name of a flower: Dianthus superbus)', /i˨rogaꜜmi/ [i˨ro˨ga˥mi˨] 'colour paper' (low tone, accent on third syllable). The system is more tonal than that of Tokyo though: there is no phonemic location for the accent unless the word has four syllables or more, so in words with 1-3 syllables the accent is either present (+) or absent (-).

Afterwards, I think the closest languages would be a number of Bantu languages, which, like Tokyo Japanese, tend to select one syllable in a word to bear the accent, which then may alter the pitch of the whole word and attached clitics or adpositions. Like Japanese, words are often phonemically accentless, gaining a default phonetic pitch pattern. Unlike Japanese, some of the languages exhibit phonemic tone in the accented syllable, and it is possible for more than one accent to appear in a word by appearing in different syllables, e.g. the name of Uganda's capital in the Luganda language, Kampala /ka˥mpala:˥˩/ [kam˥mpa˥la:˥˩] (the syllable -pa- doesn't bear phonemic accent+tone, but it gains a phonetic high pitch).

(This Luganda example may make it look like it's simply tonal, but that language has a Japanese-like rising default pitch when a word is accentless: /ebitabo/ [e˨bi˥ta˥bo˥] 'books', with phonetic low-high-high-high; and when unaccented syllables are affected by a syllable nearby with accent+tone, they simply become either high or low in predictable ways: /eddwa˥˩liro/ [e˨ddwa˥˩li˥ro˥] 'hospital', with phonetic low-falling-high-high.)

At a more removed place, I'd say you could look at BCSM (Serbo-Croatian), Swedish and Attic Greek. These languages basically combine stress and tone, and don't really have accentless words (except for many function words, like any other languages with phonemic stress), but are interesting for relevant reasons. BCSM is interesting because in the dialects that have tone in short vowels, the tones are spread onto the next syllable as well (as if the short vowel can't bear them alone, just like the pitch-accent of Japanese monosyllabic words!). And Swedish always spreads its tones (only two) throughout a word, with patterns affected by stress. It isn't really clear, but the use of the grave and acute orthographic accents in Attic Greek seems to suggest that words with alleged final stress (on the last syllable) and acute tone may have simply been accentless words in their default pitch. Attic Greek can also bear more than one accent+tone inside a single word as an effect of certain function words.

I don't know about Korean, but from the occasional short mentions in passim I've seen of it, it looks like the "pitch accent" of some dialects and Middle Korean was really just tone, spread in various patterns throughout a word (as in certain non-Bantu Niger-Congo languages like famous Yoruba). I would say those Korean varieties are simply tonal, but I might be wrong about this though (I have really read almost nothing). Lithuanian, often said to have "pitch accent", also seems irrelevant, in the sense that it doesn't have accentless words, nor tones that spread in patterns. Its two-tone distinction only appears in the very limited context of stressed heavy syllables with a long vowel or a sonorant coda, so I'd rather say it's a language with stress with a little bit of limited tone. I have no idea about Sanskrit.

2. Are there particular resources you'd recommend for learning about such?

I'd say the best thing you could do is read specialized works on relevant languages. You can find some useful surveys in general books on phonology though, like this chapter by van der Hulst from The Blackwell Companion to Phonology (2011). Surveys are useful for their discussions and classifications of many different tonal/accentual phenomena, but specialized works are also good because surveys always simplify what languages are actually like.

As some random advice, be especially cautious of Bantu literature, because their conventions are often different. To give you one example, this paper entitled "Xhosa: A tone or pitch-accent language?", discusses whether Xhosa has a "tonal pitch accent" or a "metrical pitch accent". In general linguistics, this would be translated as a discussion on whether Xhosa has a phonemic pitch accent or a phonetic (non-phonemic) pitch accent.

Similarly, someone on English Wikipedia called Kanjuzi has been adding a lot of stuff on the accent+tone systems of Luganda and Chichewa, but note the particular use of formatting (explained nowhere in the articles!): bold letters are used for the location of phonemic accent, acútes/gràves/cîrcumflexes/brĕves are used for the realizations of pitch at the phonetic level, and underlining is used to direct the reader's attention at something. Par for the course in Bantu literature.

Also, though this may be unnecessary to say, don't get too bogged down on what is or is not "tone" or "pitch accent" or "stress". In the end, in real languages you can find all sorts of combinations of suprasegmental phenomena. For example, the Cantonese tone #4 often has creaky phonation, Mandarin spreads its tones onto phonemically toneless syllables (/tjan˥˩tsi/ [tjɛn˥˩dz̩˩] 'electricity', /χə˧˥tsi/ [χɤ˧˥dz̩˧] 'box'), Turkish has been noticed to have some non-phonemic pitch patterns (even though it barely has some words with non-predictable, phonemic stress; in the article I linked to above Xhosa is described as possibly being like this too), words in Persian in some constructions may lose stress entirely in favour of another word later on (presumably French went through a stage like that before gaining the non-phonemic phrase-final stress it has today: pour voir ses deux enfants [puʁvwaʁsedøzɒ̃ˈfɒ̃] vs. Spanish para ver sus dos niños [paɾaˈβeɾ suzdoˈsixos])...

Spec Qs:
3. What are the effects of certain sounds on pitch? For instance, would velars/uvulars raise/lower pitch? What about the effect, instead of POA, of ejectivity, voicing, etc.?

I don't know about POA affecting pitch, but across time, voiced MOA very often creates lower phonemic tones (although some examples of the opposite can be found).

I dont think the consonants would interact with the pitch of the syllables. I dont really have anything else to add but just wanted to get rid of this idea up front .... of all the natlangs we have sound changes for, Im not aware of even one in which tone interacts with consonants.

I mentioned such a change in this post a month ago:
http://verduria.org/viewtopic.php?f=3&t ... 297#p22297

The Middle Chinese voiced consonants become voiceless aspirated before the level (平 píng) tone, but voiceless unaspirated before other tones.

[Nortaneous]

What do you mean by "pitch accent"?

There's a spectrum from "tone doesn't exist" to "every syllable in a word is specified for tone". It's hard to describe that spectrum, and there are more possibilities within it than you might think. Tone is more theory-dense than consonant and vowel structure -- you can make a consonant inventory by just looking at some consonant inventories, but unless you're making a language where every syllable is specified for tone and tone sandhi is rare to nonexistent, you're probably going to need some theory to explain how to get from the underlying forms to the realizations.

One way to think about "pitch accent" is that it's a word-tone system where there's some distinguishing feature that can only appear on one syllable or mora. What is that thing? What other effects (e.g. word-edge [Tinputz] or phrase-edge [diachronically in Burmese, where IIRC verbs and adjectives underwent a tone split due to details of word order and phrase-final effects]) influence the phonetic realization of the tonal contour?

These papers might help:
https://www.mcgill.ca/mcgwpl/files/mcgwpl/hyman2012.pdf
http://linguistics.berkeley.edu/~hyman/ ... accent.pdf

[Richard W]

I dont think the consonants would interact with the pitch of the syllables. I dont really have anything else to add but just wanted to get rid of this idea up front .... of all the natlangs we have sound changes for, Im not aware of even one in which tone interacts with consonants.

As Vijay has alluded to, consonants can certainly affect tone, and have in the past have led to the creation of tone and to the splits of tones. Oddly, I can't think of any cases where they have directly led to the merger of tones. As the attributes of a tone can include length, they can lead to effects on the vowels at least.

In Thai, for example, consonants even affect tone synchronically, and no syllable with an obstruent in coda position can have mid tone.

I'm a bit uncomfortable with that statement. It's certainly true that the dead and live syllables select from two different sets of tones, but I wouldn't be totally taken aback to come across tone neutralisation applied without glottal stop deletion. What has been measured is an effect of consonant voicing on the absolute pitches of tones.

[Pabappa]

Sorry, I had forgotten. There's a probable example of tonogenesis arising from coda consonants in proto-Tai, as well. But still, I don't think POA specifically would come into play here, since it would result in an odd distribution of tones, that most likely would never become truly phonemic.

[Travis B.]

Sorry, I had forgotten. There's a probable example of tonogenesis arising from coda consonants in proto-Tai, as well. But still, I don't think POA specifically would come into play here, since it would result in an odd distribution of tones, that most likely would never become truly phonemic.

Phonation and glottalization (often associated with coda plosives) readily influence tonogenesis, though, it should be noted.

[Vijay]

In Thai, for example, consonants even affect tone synchronically, and no syllable with an obstruent in coda position can have mid tone.

I'm a bit uncomfortable with that statement. It's certainly true that the dead and live syllables select from two different sets of tones, but I wouldn't be totally taken aback to come across tone neutralisation applied without glottal stop deletion. What has been measured is an effect of consonant voicing on the absolute pitches of tones.

What kind of tone neutralization do you have in mind? I'm having trouble following what your objection to what I said is.

[Richard W]

Sorry, I had forgotten. There's a probable example of tonogenesis arising from coda consonants in proto-Tai, as well. But still, I don't think POA specifically would come into play here, since it would result in an odd distribution of tones, that most likely would never become truly phonemic.

Tai languages have some statistically odd distributions of tones and segmental phonemes even in unchecked syllables, and the tone splitting added yet more markedness. In Siamese, the falling tone on short checked syllables is so rare that some reputable linguists have wrongly denied its existence.

In Phnom Penh Khmer, /r/ immediately following the onset consonant has become a tone; this immediately yields a marked distribution.

[Richard W]

What kind of tone neutralization do you have in mind? I'm having trouble following what your objection to what I said is.

Nominally low tone unstressed syllables with short vowel 'underlyingly' ending in a glottal stop are normally pronounced with the mid tone, and the glottal stop is dropped. Nominally high tone unstressed syllables with the short vowel /a/ 'underlyingly' ending in a glottal stop are normally pronounced with the mid tone, and the glottal stop dropped. I have a Thai-English dictionary (written by a Thai) that marks tones explicitly - it marks no tone for these syllables.

Examples are นคร 'city' (careful /naʔ˥ kʰɔːn˧/, casual /na˧ kʰɔn˧/) and อะไร (careful /ʔaʔ˨ rai˧/, casual /ʔa˧ rai˧/) 'what?'.

All it takes to falsify your statement is for the syllable-final glottal stop not to be dropped.

[Richard W]

In Thai, for example, consonants even affect tone synchronically, and no syllable with an obstruent in coda position can have mid tone.

Actually, I think I have a clear counter-example - มันส์ 'to enjoy; to be fun' when pronounced /mans˧/, as recorded in that Wiktionary entry. Personally, I blame that pronunciation on the teaching of English. There are a great many loanwords where a dropped final obstruent has the effect of causing the word to be in the high tone. But this isn't a loanword. The author of the Wiktionary entry is pretty clearly Thai.

[Vijay]

Interesting stuff - I didn't realize นคร was analyzed as having a glottal stop, thanks!

/mans˧/

[jal]

At a more removed place, I'd say you could look at BCSM (Serbo-Croatian), Swedish and Attic Greek. These languages basically combine stress and tone

That's an interesting take, as I've always learned there's a near universal stating that tone and stress don't go together, with very few exceptions (Papiamentu being a particular notable one).


JAL

[Estav]

I've always learned there's a near universal stating that tone and stress don't go together, with very few exceptions (Papiamentu being a particular notable one).

Do you mean "don't go together" as in "languages with contrastive tone systems don't have contrastive stress systems, and vice versa" or "the tone system of a language does not interact with that language's stress system"? Neither seems to be true in general, but I can imagine someone proposing either of these.

Standard Chinese, one of the best-known languages with a contrastive tone system, has been argued to have a contrast between stressed and unstressed syllables, although it's disputed. San Duanmu presents arguments for in "Tone and Non-tone Languages: An Alternative to Language Typology and Parameters", 2004, section 3.2.2.1, pages 902-906. It does seem to be the case that, as in many languages, stress in Chinese is predictable (I'm not sure how completely) from the internal structure of a word/phrase.

[Nerulent]

Pitch accent can describe a few different things (as I understand it):

• On one end, a stress accent system where the main distinguishing feature of stressed syllables is pitch rather volume or length, like Nubi or Ainu or Seneca - this is just a (pitch-based) stress accent system, and calling it a pitch accent system is just confusing
• Then there's systems like Serbo-Croatian, Ancient Greek and Kurtöp, with different tones possible, but still strictly one tone per word - this is the only one that I would call a pitch accent system
• On the other, there's a system like Japanese, which, depending on the dialect, is more like the tonal system of some Bantu languages, especially in that pitch marking is non-obligatory.

To me, the first two are best thought of as subtypes of an accent system - a system which has one and only one accent per word, and can be split into stress (with a further subtype of pitch-based stress) and pitch accent.

Japanese, Somali and several Bantu languages instead fall under the umbrella of 'restricted tone'. There's different subtypes in here as well, with Standard Japanese and Somali on the more restricted end, Middle Japanese and several Japanese dialects in the middle, and Luganda on the less restricted end.

Most of this is taken from pages 14-20 of this article. I'll have to read through the Hyman paper that Nort linked - it seems to go a lot more in depth and possibly come to different conclusions.

That's an interesting take, as I've always learned there's a near universal stating that tone and stress don't go together, with very few exceptions (Papiamentu being a particular notable one).

Apparently the Austronesian language Mayá also has both tone and stress (or, I guess, combines stress-accent and pitch-accent) - only the final syllable carries tone, either high or falling, but either the penult or the ultima syllables can carry stress, which is independent of the tone. Unfortunately I don't have any examples.

[Nortaneous]

Japanese, Somali and several Bantu languages instead fall under the umbrella of 'restricted tone'. There's different subtypes in here as well, with Standard Japanese and Somali on the more restricted end, Middle Japanese and several Japanese dialects in the middle, and Luganda on the less restricted end.

Restricted tone as in a tonal system best analyzed as certain tones contrasting with tonelessness, where full, stress-bearing lexical words are permitted to be toneless? (The name isn't great; most tonal systems have some sorts of restrictions. That's one of the most important things for conlangers to understand about tone: you probably want to have restrictions on the structure of tone, even if they're as simple as Mandarin tone sandhi.)

That's an interesting take, as I've always learned there's a near universal stating that tone and stress don't go together, with very few exceptions (Papiamentu being a particular notable one).

Possibly also Yonghe Qiang.

[Moose-tache]

Pitch accent can describe a few different things (as I understand it):

• On one end, a stress accent system where the main distinguishing feature of stressed syllables is pitch rather volume or length, like Nubi or Ainu or Seneca - this is just a (pitch-based) stress accent system, and calling it a pitch accent system is just confusing
• Then there's systems like Serbo-Croatian, Ancient Greek and Kurtöp, with different tones possible, but still strictly one tone per word - this is the only one that I would call a pitch accent system
• On the other, there's a system like Japanese, which, depending on the dialect, is more like the tonal system of some Bantu languages, especially in that pitch marking is non-obligatory.

I’m struggling to see much practical difference among these systems. The de Boer article you linked to describes “restricted tone” in a way that equally applies to Mohawk, and other languages that you would call “pitch-based stress accent” systems. You can conceptualize them any number of ways, but the result is the same. Here, let's try.
Theory 1: Tone melodies – tone is a property of words, not syllables. The language has a number of tonal contours that a word may belong to, such as “down, down, up” or “falling-rising” or whatever.
Theory 2: Stress that’s totally not a tone – each word has a stressed syllable, and it just happens to involve a change in resonant frequency, but it’s not really a tone.
Theory 3: Tone that’s totally not stress – each word has a “high” toneme on one syllable, but it’s only a coincidence that they behave exactly like stressed syllables; it’s just a single restricted toneme you guys.
Theory 4: Stress melodies – there is one stressed syllable, but it can be “stressed” in different ways, such as a high tone or a sharply falling tone; it’s tone or stress depending on how we’re feeling today. Also, as the name suggests these can just be called tone melodies if you want to bring it full circle.

The boundaries between stress and pitch-accent are equally arbitrary (how much volume do we need to have a volume-based stress system?). And it's not like fully tonal systems treat each syllable as independent entities (enough sandhi would reduce even Cantonese to a pitch-accent system in which only the conditioning syllable has any remaining phonemic tone information). I think I'm kind of proving your point though, in a way, because the takeaway here is that we need to talk about specific mechanisms not archetypes. There isn't going to be one official kind of pitch accent system anymore than there is one official kind of person-suffix paradigm.

[Nerulent]

Restricted tone as in a tonal system best analyzed as certain tones contrasting with tonelessness, where full, stress-bearing lexical words are permitted to be toneless? (The name isn't great; most tonal systems have some sorts of restrictions. That's one of the most important things for conlangers to understand about tone: you probably want to have restrictions on the structure of tone, even if they're as simple as Mandarin tone sandhi.)

Well it makes the most sense for systems like Tokyo Japanese and Somali, where there is up to one tone bearing unit in a word. That's a hell of a restriction I'll try get back to after reading through those Hyman papers.

I’m struggling to see much practical difference among these systems. The de Boer article you linked to describes “restricted tone” in a way that equally applies to Mohawk, and other languages that you would call “pitch-based stress accent” systems.

I'm not exactly sure what you mean by practical differences - there are obviously differences between the systems. Do you mean the distinctions aren't useful? My aim is to give the term pitch-accent a more specific, coherent meaning (i.e. to describe the similarities in the prosodic systems of Greek, Serbo-Croatian etc.). As the papers linked in this thread show, the term pitch-accent has been used for anything in between a full stress-accent system and a full tone system, rendering it kinda useless.

Mohawk has an obligatory, culminative accent - this gives it a similar system to Ainu or Nubi or most of the other Iroquois languages, and it differs from Somali and Tokyo Japanese where the accent/tone is not obligatory. Yes, these systems are similar, but the Mohawk system is also similar to a stress-accent language like English in a way that Japanese isn't. (Mohawk and Iroquois have the complication that there are multiple surface tones, but underlyingly there is only a single kind of stress).

You can conceptualize them any number of ways, but the result is the same. Here, let's try.
Theory 1: Tone melodies – tone is a property of words, not syllables. The language has a number of tonal contours that a word may belong to, such as “down, down, up” or “falling-rising” or whatever.
Theory 2: Stress that’s totally not a tone – each word has a stressed syllable, and it just happens to involve a change in resonant frequency, but it’s not really a tone.
Theory 3: Tone that’s totally not stress – each word has a “high” toneme on one syllable, but it’s only a coincidence that they behave exactly like stressed syllables; it’s just a single restricted toneme you guys.
Theory 4: Stress melodies – there is one stressed syllable, but it can be “stressed” in different ways, such as a high tone or a sharply falling tone; it’s tone or stress depending on how we’re feeling today. Also, as the name suggests these can just be called tone melodies if you want to bring it full circle.

The boundaries between stress and pitch-accent are equally arbitrary (how much volume do we need to have a volume-based stress system?). And it's not like fully tonal systems treat each syllable as independent entities (enough sandhi would reduce even Cantonese to a pitch-accent system in which only the conditioning syllable has any remaining phonemic tone information). I think I'm kind of proving your point though, in a way, because the takeaway here is that we need to talk about specific mechanisms not archetypes. There isn't going to be one official kind of pitch accent system anymore than there is one official kind of person-suffix paradigm.

Are these theories meant to be applied to what I'm calling a pitch-based stress accent system? Because theories 1 and 4 don't really fit - 1 is describing what I'm calling a restricted tone system, and 4 is what I'm calling pitch-accent. 2 and 3 are sort of complementary - that's why I'm trying to differentiate accent from stress.

Edit: I'm not married to the terms I'm using, just the distinctions. Feel free to offer up some better terms

[Richard W]

Standard Chinese, one of the best-known languages with a contrastive tone system, has been argued to have a contrast between stressed and unstressed syllables, although it's disputed. San Duanmu presents arguments for in "Tone and Non-tone Languages: An Alternative to Language Typology and Parameters", 2004, section 3.2.2.1, pages 902-906. It does seem to be the case that, as in many languages, stress in Chinese is predictable (I'm not sure how completely) from the internal structure of a word/phrase.

Thai, though more monosyllabic than English, is widely described as having iambic stress. Toneless syllables are fairly restricted, and I think not universally acknowledged.
.