Linguistic Miscellany Thread

[zompist]

What I don't understand is how exactly are Chinese rhymes and "classical phonemes" different? What is it that one of them has and the other lacks? As far as this scheme is concerned, tiān is two phonemes and a tone, so phonemicity is well-defined.

If by "phoneme" you mean "any sequence of sounds whatsoever", then the concept has been loosened so far as to be effectively meaningless. Just call this one a rhyme and don't pretend that it's the same as traditional phonemes.

(Also, FWIW, you can create a traditional phonemic analysis for Mandarin. Jerry Norman's book has one. It just ends up seeming more rickety and jerry-rigged than one for English. Kuchigakatai's post is just the beginning!)

Another question: say that you're making tables of a language's paradigms, e.g.:

NOM -ok
ACC -en
GEN -in
LOC -ut

But then you stumble upon a noun class that doesn't mark its nominative and doesn't have a locative. How would you denote that, if you had to have a row for the locative? (sorry for the contrived example, the question is for the general case).

Grammars I've seen usually leave out forms that don't exist, and often merge categories (e.g. if ACC and GEN are the same in the plural, you label that row GEN/ACC). For natlangs, there are usually so many paradigms you have to list that you welcome any chance to make them simpler.

Naturally, you want to explain what you're doing in the text— e.g. mention up front that not all locatives exist, and say what happens then.

Oh, and a zero morph is of course different from no morph at all. So if the nominative exists but has no suffix, you leave it in the table. If you used hyphens for the other cases, a naked hyphen will be clear enough.

[Richard W]

It is possible to map the rhymes into phonemes of vowel nucleus and consonants, but the rhymes have such complementary distributions and allophony that it is pretty ambiguous how to go about it.

There is no necessity for there to be a single correct analysis. It is entirely conceivable that different speakers are working with different analyses. Furthermore, there is no guarantee that all instances of allophones are recognised as such, and some speakers may even phonemicise what are generally recognised as allophonic differences.

[Travis B.]

It is possible to map the rhymes into phonemes of vowel nucleus and consonants, but the rhymes have such complementary distributions and allophony that it is pretty ambiguous how to go about it.

There is no necessity for there to be a single correct analysis. It is entirely conceivable that different speakers are working with different analyses. Furthermore, there is no guarantee that all instances of allophones are recognised as such, and some speakers may even phonemicise what are generally recognised as allophonic differences.

This is my thought. There is no reason for individual speakers' internal analyses to map onto the ideal analyses linguists come up with.

[Nortaneous]

Yes, it has obvious constituent parts in its realisation, but we don't analyse English /eɪ/ as two phonemes, do we?

Well, some people don't. But even aside from American /ej/, I've seen plenty of analyses that take diphthongs as two phonemes packed into one nucleus, usually for Austronesian and Papuan languages that have a lot of diphthongs.

But is "diphthong" a single pattern? Sometimes diphthongs fill slots in the vowel grid:

Code: Select all

Mid-Atlantic English with TRAP-BATH split
i  ɚ  u
eə    oə
ɛ     ʌ
æ     ɑ

Romanian
i  ɨ u
e  ə o
ea a oa

Chechen
i  y  u
ie yø uo
e  ø  o
æ     ɑ

Adzera
i  u
ai au
  a
  
Chong (simplified, ignoring phonation and length)
i  ɤ  u 
iə ɤə uə
e  ə  o 
ɛ  ʌ  ɔ   
ai a  ao      

And sometimes they don't:

Code: Select all

Aneme Wake (and many other languages)
i u
e o  ei    oi  eu    ou
 a      ai        au	

Fasu
i u
e o
 a   ae  ao  ai  au

Gizrra
i ɨ u
e ə o
  a
ae ao ai aɨ au
əe
ɨe    ɨi
      ui

Iau
i ɨ u
ɪ   ʊ
e   o
  a
ae aɪ aʊ ai aɨ au
         ei
oe       oi
   ʊɪ
         ui
aʊɪ aui

[Travis B.]

Yes, it has obvious constituent parts in its realisation, but we don't analyse English /eɪ/ as two phonemes, do we?

Well, some people don't. But even aside from American /ej/, I've seen plenty of analyses that take diphthongs as two phonemes packed into one nucleus, usually for Austronesian and Papuan languages that have a lot of diphthongs.

But is "diphthong" a single pattern? Sometimes diphthongs fill slots in the vowel grid:

And then there is the case of where there are phonemic diphthongs, allophonic diphthongs, and diphthongs derived from vowels plus vocalized consonants. Take (for the sake of familiarity), the English here:

There are four phonemic diphthongs: /ae əe ɑɔ ɔɪ/ (note that the contrast between /ae əe/ is relatively marginal).

But at the same time there are the following allophonic diphthongs derived from single segments: [ɛə~eə eɪ ou] from /ɛ e o/ respectively.

But at the same time there are the following diphthongs derived from /Vl/: [aɤ ɒo ɛɤ ɜɤ ʌɤ eɯ ɘɯ oʊ iɯ uʊ] from /al ɒl ɛl ɜl ʌl el ɘl ol il ul/ respectively.

[Kuchigakatai]

There is no necessity for there to be a single correct analysis. It is entirely conceivable that different speakers are working with different analyses. Furthermore, there is no guarantee that all instances of allophones are recognised as such, and some speakers may even phonemicise what are generally recognised as allophonic differences.

That is an excellent point and one I needed to hear...

[Kuchigakatai]

Do you guys know of languages with a fairly pure ergative-absolutive alignment? As in, maybe they have another alignment too, but it's used only in a peculiar minor situation or two like purpose clauses when done with a specific subordinator, or something.



I was also just thinking that the various nominative-accusative languages whose grammars I'm familiar with are also split-accusative...
- English and French have direct alignment in action nouns ("the firing of the CEO" <- did the CEO fire someone, or was s/he fired?)

- Spanish has it quite a bit with inanimates in both regular finite clauses (via fronting with focus) and infinitive clauses ("al acabar la tarea" <- did the task end, or was the task finished?, cf. al acabar mi amigo la tarea 'right after my friend finished the task') besides also action nouns

- Standard Arabic has it in verbal nouns too (ʔaklu l-kalbi <- did the dog eat something or did someone eat the dog?), which is more notable than English and French as they're used a bit more often, in more constructions

- Mandarin also gets in this list with its ambiguous use of topicalization of direct objects—a classic example in Chinese linguistics is 魚還沒吃啊 yú hái méi chī a = 'the fish haven't eaten [food] yet' ~ 'you/someone/we/etc. haven't eaten the fish yet', and verbs inside relative clauses can present some ambiguity about their arguments

I'm starting to feel the term split-X is only useful when a major sentence type involves another alignment, like all finite clauses in the past having another alignment. There might not be languages with a pure use of alignments...

[Creyeditor]

Have you read "the blue bird of ergativity"? You will find it on google and IMHO it adresses the issue quite clearly.

[Kuchigakatai]

Have you read "the blue bird of ergativity"? You will find it on google and IMHO it adresses the issue quite clearly.

No, first time I hear of it! Thanks.

[bradrn]

Do you guys know of languages with a fairly pure ergative-absolutive alignment? As in, maybe they have another alignment too, but it's used only in a peculiar minor situation or two like purpose clauses when done with a specific subordinator, or something.

I know it’s been reported for one Australian language, but I can’t seem to find it now.

I was also just thinking that the various nominative-accusative languages whose grammars I'm familiar with are also split-accusative...
- English and French have direct alignment in action nouns ("the firing of the CEO" <- did the CEO fire someone, or was s/he fired?)

- Spanish has it quite a bit with inanimates in both regular finite clauses (via fronting with focus) and infinitive clauses ("al acabar la tarea" <- did the task end, or was the task finished?, cf. al acabar mi amigo la tarea 'right after my friend finished the task') besides also action nouns

- Standard Arabic has it in verbal nouns too (ʔaklu l-kalbi <- did the dog eat something or did someone eat the dog?), which is more notable than English and French as they're used a bit more often, in more constructions

- Mandarin also gets in this list with its ambiguous use of topicalization of direct objects—a classic example in Chinese linguistics is 魚還沒吃啊 yú hái méi chī a = 'the fish haven't eaten [food] yet' ~ 'you/someone/we/etc. haven't eaten the fish yet', and verbs inside relative clauses can present some ambiguity about their arguments

I'm starting to feel the term split-X is only useful when a major sentence type involves another alignment, like all finite clauses in the past having another alignment. There might not be languages with a pure use of alignments...

I’m also gradually coming to the conclusion that ‘split alginment’ is a somewhat meaningless phrase — there’s just too much variation between split types, and no particular defining features which separates ‘split’ from ‘non-split’ alignment. (That reminds me, I’ve been meaning to write an expanded version of my ergativity thread covering this stuff… I’d better get started when I have more time.)

That being said, I’m not entirely sure your examples are actually split systems in the same way that, say, Dyirbal and Tibetan are. Firstly, action nouns are really something which takes place within the NP — they’ll always be somewhat different to case-marking and other morphology. Most languages have funny things going on with nominalisations, so I’m not sure it’s useful to call case-marking in nominalisations ‘split’. (Though English does have a split system, in the form of direct marking for nouns vs accusative for pronouns.) I can’t comment on the Mandarin example.

Have you read "the blue bird of ergativity"? You will find it on google and IMHO it adresses the issue quite clearly.

Yep, it’s one of my favourite articles! You can read my response to it here.

[Richard W]

Most languages have funny things going on with nominalisations, so I’m not sure it’s useful to call case-marking in nominalisations ‘split’. (Though English does have a split system, in the form of direct marking for nouns vs accusative for pronouns.)

Would you please dejargonise that. It doesn't look right, but I can't be sure until I understand it.

[Kuchigakatai]

That being said, I’m not entirely sure your examples are actually split systems in the same way that, say, Dyirbal and Tibetan are. Firstly, action nouns are really something which takes place within the NP — they’ll always be somewhat different to case-marking and other morphology. Most languages have funny things going on with nominalisations, so I’m not sure it’s useful to call case-marking in nominalisations ‘split’. (Though English does have a split system, in the form of direct marking for nouns vs accusative for pronouns.)

I feel you're getting tricked/confused by the terminology there though. English also has a nominative-accusative system in nouns. The definitions of morphosyntactic alignments involve syntax too, they're not just about morphology:

My friend sings.
My friend sings the anthem.

In both sentences, the subject (S and A) is marked with word order, appearing before the verb, while O appears afterwards. English thus in basic sentences shows nom.-acc. alignment in nouns.

Similarly, in action nouns you could have S=A but not =O, in the way things are marked via word order or prepositions or cases or marking within the action noun, etc. But English/French/Spanish/Arabic instead have ambiguous marking, where S=A=O ("the firing of the CEO").

Salmoneus once proposed a conlang idea I found very amusing, with a direct-inverse alignment accomplished purely via syntax, by requiring uninflected auxiliary verbs when this or that happened in the animacy hierarchy...

[bradrn]

Most languages have funny things going on with nominalisations, so I’m not sure it’s useful to call case-marking in nominalisations ‘split’. (Though English does have a split system, in the form of direct marking for nouns vs accusative for pronouns.)

Would you please dejargonise that. It doesn't look right, but I can't be sure until I understand it.

Sorry, which jargon do you want me to explicate? I don’t see anything particularly confusing there myself.

That being said, I’m not entirely sure your examples are actually split systems in the same way that, say, Dyirbal and Tibetan are. Firstly, action nouns are really something which takes place within the NP — they’ll always be somewhat different to case-marking and other morphology. Most languages have funny things going on with nominalisations, so I’m not sure it’s useful to call case-marking in nominalisations ‘split’. (Though English does have a split system, in the form of direct marking for nouns vs accusative for pronouns.)

I feel you're getting tricked/confused by the terminology there though. English also has a nominative-accusative system in nouns. The definitions of morphosyntactic alignments involve syntax too, they're not just about morphology: …

This is why it’s useful to carefully separate the various areas where morphosyntactic alignment can appear! Specifically, for English:

• Verbal agreement is accusative for 3SG, direct/absent for everything else (which interestingly violates the animacy hierarchy!)
• Case-marking is accusative for pronouns, direct/absent for everything else
• Word order is consistently accusative.

Of course, in English, everything turns out either accusative or direct, and everything is accusative at least in word order, but it’s interesting to note that the conditioning environment changes depending on the area, and of course there are many languages with completely different alignments in each place. (So for instance, you get ones like Komnzo, with active-stative agreement, ergative case-marking, and neutral word order.)

Similarly, in action nouns you could have S=A but not =O, in the way things are marked via word order or prepositions or cases or marking within the action noun, etc. But English/French/Spanish/Arabic instead have ambiguous marking, where S=A=O ("the firing of the CEO").

…yes? I mean, I agree with all of this, I just don’t understand what your point is here.

Salmoneus once proposed a conlang idea I found very amusing, with a direct-inverse alignment accomplished purely via syntax, by requiring uninflected auxiliary verbs when this or that happened in the animacy hierarchy...

It gets weirder! Movima has direct-inverse word order. Not only that, it’s fully syntactically direct-inverse as well — so, for instance, you can relativise on the intransitive subject, or on the less animate transitive argument, but not on the more animate transitive argument.

[Kuchigakatai]

This is why it’s useful to carefully separate the various areas where morphosyntactic alignment can appear! Specifically, for English:

• Verbal agreement is accusative for 3SG, direct/absent for everything else (which interestingly violates the animacy hierarchy!)
• Case-marking is accusative for pronouns, direct/absent for everything else
• Word order is consistently accusative.

Of course, in English, everything turns out either accusative or direct, and everything is accusative at least in word order, but it’s interesting to note that the conditioning environment changes depending on the area, and of course there are many languages with completely different alignments in each place. (So for instance, you get ones like Komnzo, with active-stative agreement, ergative case-marking, and neutral word order.)

That seems nice and fair, and it's definitely interesting, but AFAIK your use of the jargon is also non-standard? I thought that if any of those distinctions was present, to distinguish A from O, aligning the former with S, this would mean saying e.g. "English has ... direct marking for nouns" would be unusual (or confusing/confused use of the terminology). My point in that post was to direct attention away from morphological case marking.

It gets weirder! Movima has direct-inverse word order. Not only that, it’s fully syntactically direct-inverse as well — so, for instance, you can relativise on the intransitive subject, or on the less animate transitive argument, but not on the more animate transitive argument.

Beautiful!

I sometimes think of transitive subjects (A, which is typically animate) as "less core" than transitive objects (O). Something to do with where focus usually falls in when a transitive verb is used (I have the intuitive feeling "the princess ate the apple" would typically have more focus on the apple, even though both NPs are definite), also, all the common basic word orders are subject->object: SVO, SOV, VSO... which I feel says something about the statistical importance of O. A language that is not comfortable relativizing the more animate transitive argument sounds very interesting for the same reason. Hmm...

Also, heh, have you seen the phrase "to relativize on [a constituent]" before? I used to say that all the time, but people here (esp. Salmoneus) complained it was inappropriate...... I think people told me to just say "to relativize [a constituent]".

[bradrn]

This is why it’s useful to carefully separate the various areas where morphosyntactic alignment can appear! Specifically, for English:

• Verbal agreement is accusative for 3SG, direct/absent for everything else (which interestingly violates the animacy hierarchy!)
• Case-marking is accusative for pronouns, direct/absent for everything else
• Word order is consistently accusative.

Of course, in English, everything turns out either accusative or direct, and everything is accusative at least in word order, but it’s interesting to note that the conditioning environment changes depending on the area, and of course there are many languages with completely different alignments in each place. (So for instance, you get ones like Komnzo, with active-stative agreement, ergative case-marking, and neutral word order.)

That seems nice and fair, and it's definitely interesting, but AFAIK your use of the jargon is also non-standard? I thought that if any of those distinctions was present, to distinguish A from O, aligning the former with S, this would mean saying e.g. "English has ... direct marking for nouns" would be unusual (or confusing/confused use of the terminology). My point in that post was to direct attention away from morphological case marking.

I’m not sure whether my use is standard or not. On the other hand, I certainly find it a lot clearer than simply saying ‘English nouns use accusative marking’, since that lumps together all the different ways in which accusative marking can occur. (You wouldn’t be able to get away with saying this in a language like Komnzo.)

It gets weirder! Movima has direct-inverse word order. Not only that, it’s fully syntactically direct-inverse as well — so, for instance, you can relativise on the intransitive subject, or on the less animate transitive argument, but not on the more animate transitive argument.

Beautiful!

I sometimes think of transitive subjects (A, which is typically animate) as "less core" than transitive objects (O). Something to do with where focus usually falls in when a transitive verb is used (I have the intuitive feeling "the princess ate the apple" would typically have more focus on the apple, even though both NPs are definite), also, all the common basic word orders are subject->object: SVO, SOV, VSO... which I feel says something about the statistical importance of O. A language that is not comfortable relativizing the more animate transitive argument sounds very interesting for the same reason. Hmm...

There’s also the fact that VO is often analysed as forming a constituent. Except that Movima’s the opposite — if anything, it looks like the verb and the transitive subject form a constituent, with O/S being ‘added on’ to that core bit. It’s certainly unusual in the scheme of things.

Also, heh, have you seen the phrase "to relativize on [a constituent]" before? I used to say that all the time, but people here (esp. Salmoneus) complained it was inappropriate...... I think people told me to just say "to relativize [a constituent]".

I wouldn’t really know… I’m not a syntactician. (Not even close!)

[KathTheDragon]

Given the use of "direct" in direct/inverse, it would be better to refer to "no distinctions" as "neutral" marking.

[Richard W]

Verbal agreement[/b] is accusative for 3SG, direct/absent for everything else (which interestingly violates the animacy hierarchy!)

So We drink water shows accusative verbal agreement, but We eat peas has direct or absent verbal agreement?

Case-marking[/b] is accusative for pronouns, direct/absent for everything else

Except that English is marked nominative; who is the only pronoun for which the nominative seems to be the marked form, and many speakers lack whom.

[bradrn]

Verbal agreement[/b] is accusative for 3SG, direct/absent for everything else (which interestingly violates the animacy hierarchy!)

So We drink water shows accusative verbal agreement, but We eat peas has direct or absent verbal agreement?

Not quite. We and peas show direct/absent verbal agreement, while water shows accusative agreement:

We drink water
Water drink-s us
Peas drink us

Case-marking[/b] is accusative for pronouns, direct/absent for everything else

Except that English is marked nominative; who is the only pronoun for which the nominative seems to be the marked form, and many speakers lack whom.

This is quite correct — though often not acknowledged as such, English does indeed seem to be marked nominative.

[Richard W]

And what about

• I was him in that play
• I was tired and
• If that be true...

?

[bradrn]

And what about

• I was him in that play
• I was tired and
• If that be true...

?

What about them? Copulae are often weird with regards to alignment. Specifically, in English:

• Unlike all other verbs, the copula exhibits agreement with the number of the subject (counting you as a plural), as well as a special form for first person singular.
• Case-marking becomes a bit wonky as well: the subject is consistently in the nominative, but the object can be either nominative or accusative depending on how learnèd you want to be. (IIRC Dixon has noted that the copula taking a nominative subject and accusative object is a characteristic of marked-nominative languages.)