Message-Id: <m0lfXjx-0001vQC@snark.thyrsus.com>
From: cbmvax!snark.thyrsus.com!cowan@uunet.UU.NET (John Cowan)
Subject: Grammar by tones in natural languages
To: jimc (jim carter)
Date: Mon, 27 Apr 92 11:35:39 EDT


I posted the following message to linguist-list:

> Does anybody know of any languages in which tones are used not to discriminate
> lexical items but to specify syntactic relations?  In other words, are there
> any languages in which (to make up an example) >eat< means 'eat' and >dog<
> means 'dog' regardless of tone, but where:
> 
> 	eat1 dog1	means		'something eats and is a dog'
> 	eat4 dog6	means		'an eater of dogs'
> 	eat1 dog4	means		'the dog eats'
> 	eat4 dog1	means		'the eater is a dog'
> 
> and so on?

Note the choice of tones for the "made-up" examples.  :-)

The rest of this message is a summary of the replies I got.
It seems that -gua!spi style tones are unusual but not unknown in the world's
languages.  I would point out that the African-style tones are much simpler
than -gua!spi or Chinese, and are typically limited to high and low.

---------------

I received responses to my query about "grammar by tones" from the following:
	Tucker Childs <099CHILD@witsvma.wits.ac.za>
	Jeff Lansing <lansign@bend.ucsd.edu>
	Eric Schiller <schiller@sapir.uchicago.edu>
	Kathleen Hubbard <hubbard@garnet.berkeley.edu>
	Randy LaPolla <HSLAPOLLA@ccvax.as.edu.tw>
	John Goldsmith <gldsmth@bloomfield.uchicago.edu>
	John Kingston <KINGSTON@cs.umass.edu>
	Malcolm Ross <mdr412.coombs.anu.edu.au>
	Chet Creider <creider@cogsci.uwo.ca>
	<SABINO@ducvax.auburn.edu>
I wish to express my appreciation to all of these people for their kind
assistance.

Childs, Hubbard, Goldsmith, and Kingston cite examples from Niger-Congo,
mostly from Bantu languages.

Childs:

> In ... Sotho (Bantu), Kisi (Atlantic), it is usually verbs that lose
> lexical tone before nouns, and they are typically more heavily involved
> in grammatical tone (tense/aspect/mood distinctions).

Hubbard:

> In a lot of the Eastern Bantu languages, tone has an extremely
> low lexical load (i.e. there are only a few lexical minimal pairs
> distinguished solely by tone), and a high grammatical load
> (distinguishing one verb tense from another, etc.) I guess you'd
> say the morphological function is greater than the syntactic-
> relations function, but there are a few things: in Runyambo (NW
> Tanzania), there is a High Tone Deletion rule that applies in
> certain syntactic contexts, such as genitive phrases and verb-
> complement phrases. One thing the tonal pattern can tell you is
> whether the word following the verb is part of the same clause or
> rather a vocative or right-dislocated subject:
> 
> /babona' kato'/	-->	babona kato'	"they see Kato"	(deletion)
> vs.
> /babona' kato'/	-->	babona' kato'	"they see, Kato" (no deletion)
> /abona' kato'/	-->	abona' kato'	"he sees, Kato does" (ditto)
> 
> Another interesting deletion fact is that the rule normally does not apply
> between a noun and a following adjective, but it does when the sequence has
> become somehow "lexicalized".  For instance, the phrase "omuka'zi muku'ru"
> means 'old woman', but if the H tone on the noun is deleted, "omukazi muku'ru"
> means 'eldest wife'.  Sort of a compound.  Likewise, "embu'zi m'bi" means
> 'bad goat', but when it appears in a well-known proverb it's "embuzi m'bi"
> -- with the H tone on the noun deleted  -- as though in that context
> "bad-goat" is a lexical item.

Goldsmith:

> Igbo has quite fixed word order, so the subject will always precede  
> the verb and the object will always follow it.  But in a range of  
> syntactic constructions, there is a rule of tonal mutation that will  
> modify the tone of the object noun in much the way that you're asking  
> for; and in a smaller number of constructions, there is a floating  
> tone in between the subject and the verb which moves leftward,  
> attaching to the noun and modifying its tonal pattern also.

Kingston:

> In Bakweri, a Bantu language of the Cameroon, the tone pattern on the verb
> in a relative clause differs between cases where the head noun is
> coreferent with the subject of the relative clause and that where it's
> coreferent with the object, so tone is, indirectly, related to the
> expression of grammatical relations.

Creider mentions something even closer to what I had in mind:

> All of the Southern Nilotic and most of the Eastern Nilotic languages
> use tone to signal syntactic case. E.g. (Kipsigis, Southern Nilotic)
> 
> ke:re tE:tA (f-h l-h)  (s/he is looking at the cow)
> see   cow
> 
> ke:re tE:tA (f-h l-l)  (the cow is looking at her/him)
> 
> (Nandi, Southern Nilotic)
> ame se:se:nik (h-l h-h-f)  (it eats dogs)
> eat dogs
> 
> ame se:se:nik (h-h l-h-l)  (dogs eat it)

(Interestingly, my only available information on Nandi classifies it as
Bantu/Niger-Congo, not Nilotic.  Is this a slip, an error in my information,
a genuine dispute, or two languages with the same name?)

Schiller says that Hmong illustrates syntactic tone, but without details.
LaPolla refers to the use of tone to specify verb aspect in some Chinese
dialects.  Ross refers to a case in Takia (Austronesian) from his own research,
but says it is still tentative.  Sabino says that Virgin Islands Creole English
distinguishes future from negative by overall sentence contour, which doesn't
fit my paradigm (English distinguishes statement from ironic question by
contour, after all).

Finally, Lansing points to English:

> A _silver knife_ (low high) is a knife made out of silver, and a _silver
> knife_ (high low) is a knife for cutting silver.

-----

Informal Bibliography (apologies for the lack of proper citations):

Martha Ratliff's article on Tonal doublets in LTBA (Linguistics of the
Tibeto-Burman area).

Kathleen Hubbard, an LSA paper in January on the syntax-phonology interactions
in Runyambo.

Green and Igwe's grammar of Igbo.

John Goldsmith's 1976 Ph.D. dissertation from MIT,
available from the IULC, entitled Autosegmental Phonology.  

Orin Gensler at the linguistics department, University of California at
Berkeley; a master's thesis on Bakweri tone.

Sprauve, Gilbert. 1974. Towards a Reconstruction of Virgin Islands Creole
Phonology. Ph.D diss., Princeton University.

--
cowan@snark.thyrsus.com		...!uunet!cbmvax!snark!cowan
		e'osai ko sarji la lojban