Re: polysynthetic languages

On Fri, 19 Sep 2003 16:42:17 -0600, Dirk Elzinga <dirk_elzinga@...>
wrote:

>On Friday, September 19, 2003, at 12:09  PM, Tim May wrote:
>>
>> Incidentally, Dirk, what would _you_ consider the defining quality of
>> a polysynthetic language?  Polypersonalism?  Certain types of
>> object-incorporation?  One I read recently - an open class of bound
>> morphemes?  Definitains like "sufficiently synthetic that a sentence
>> may be a single word" have always struck me as inadequate, at least
>> without further elaboration.  After all, a single verb can form a
>> sentence in some highly isolating languages.
>>
>> I realize it may not be possible to give an absolutely final answer on
>> this question, but what principle guides your own use of the term,
>> generally speaking?
>
>Impressionistic principles guide my use of the term 'polysynthetic.'
>For me generally, if a verbal word contains content morphemes besides
>the verbal root, that is an indication of a polysynthetic nature.
>Joseph Greenberg proposed a simple way of quantifying the typological
>cast of a language. I'll summarize for two of the categories he
>discusses: (poly)synthesis and agglutination.
>
>To discover the degree of synthesis present in a language, take a
>sample text of sufficient size. For each word of the text, count the
>number of morphemes. (The term 'morpheme' is defined by Greenberg as
>the minimum meaningful sequence of phonemes in a language.) The
>synthetic index will be the average number of morphemes per word. The
>higher the number, the more synthetic the language. Greenberg gives the
>following figures for various languages:
>
>Eskimo: 3.72
>Sanskrit: 2.59
>Swahili: 2.55
>Yakut: 2.17
>Anglo-Saxon: 2.12
>English: 1.68
>Farsi: 1.52
>Vietnamese: 1.06
'Yemls: 1.61

I only have a couple texts and both are outdated (as well as being relay
texts!), but the numbers are probably close enough. One text got 1.59 and
the other 1.63. It looks like playing with word and morpheme definitions
doesn't really change things. 'Yemls turned out to be less synthetic than I
thought, I guess because words with complicated aspect combinations don't
occur all that often.

>Eskimo, which is usually held to be polysynthetic, has the highest
>synthesis index. Vietnamese, which is usally held to isolating, has the
>lowest. Greenberg proposes ranges which roughly coincide with
>impressionistic categorizations of languages:
>
>analytic: 1.00-1.99
>synthetic: 2.00-2.99
>polysynthetic: 3.00+
>
>The degree of agglutination in a language is expressed as a "ratio of
>agglutinative constructions to morph junctures". Roughly, what is
>involved is the degree of morphophonemic alternation found in an
>utterance. An agglutinative construction is one in which both morphs
>belong to morphemes which are automatic. A morpheme is automatic if its
>morphs alternate in predictable fashion, or shows no alternation at all
>(morph : morpheme :: phone : phoneme). So in the word 'leaves', there
>are two morphs /liv-/ and /z/, both of which belong to morphemes which
>alternate in completely predictable fashion (i.e., /liv-/ alternates
>with /lif/ before the plural /z/; /z/ occurs when the noun stem does
>not end in a voiceless consonant or a sibilant). If alternations are
>not automatic, the agglutinative index will go down. Here are
>Greenberg's figures for the same languages:
>
>Eskimo: 0.03
>Sanskrit: 0.09
>Anglo-Saxon: 0.11
>English: 0.30
>Farsi: 0.34
>Yakut: 0.51
>Swahili: 0.67
>Vietnamese: ...
I haven't run numbers for this, but if I understand correctly, the result
should be pretty high. 'Yemls has a fair amount of alternation, but they
are very regular.

>Vietnamese doesn't rate since there are next to no junctures within
>words. Notice that Eskimo has the lowest degree of agglutination, and
>Swahili the highest. Apparently there are more non-automatic
>morphophonemic alternations going on in Eskimo than in the other
>languages included by Greenberg. He proposes that languages which have
>an index above 0.50 be called 'agglutinative'.
>
>> (I am prompted to ask this by a recent reading of Jacques Guy's
>> postings on sci.lang.  It appears that he considers French to be
>> agglutinative but not polysynthetic, and remarks on the extreme rarity
>> of polysynthetic languages outside the Americas.  It seems to me that
>> this depends on where you draw the line, which I've never been certain
>> of.  And you're the logical person to ask.)
>
>I don't work professionally with polysynthetic languages, at least I
>don't think I do. The Uto-Aztecan languages are impressionistically
>simpler than Eskimo (for example), and the one UA language taken by
>many to be polysynthetic, Nahuatl, doesn't strike me as being
>particularly so. But I haven't run the numbers à la Greenberg to see if
>that's the case.
>
>By contrast, my constructed language Miapimoquitch is polysynthetic to
>roughly the same degree as Salishan languages are (perhaps a bit less).
>Again, I'd need to run the numbers and see.
Speaking of Miapimoquitch, is it up anywhere? Also, the Tepa page doesn't
seem to be available anymore. I suppose I could try googling in both cases.

(nameless)

>Dirk
>--
>Dirk Elzinga
>Dirk_Elzinga@byu.edu
>