Morphosyntactic alignment

The best topics from Languages & Linguistics, kept on a permanent basis.
Posts: 10
Joined: Fri Apr 29, 2011 11:31 am

Re: Morphosyntactic alignment

Post by hadad »

I'm not certain what the alignment would be called, but here's how Galleme is structured:

Nominative Agent
Irajedme tera
Ira-j-ed-me te-ra
I went (willfully) to you

Nominative Patient
Irajedmi tem
Ira-j-ed-mi te-ra
I was brought (by someone else/beyond my will) to you

Accusative -(e)m
You brought me
Irajedte mem
Ira-j-ed-te me-m

User avatar
Posts: 188
Joined: Mon Aug 24, 2009 5:23 pm

Re: Morphosyntactic alignment

Post by ná'oolkiłí »

I just finished Alice C. Harris's Diachronic Syntax: The Kartvelian Case, and the account and reconstruction of MSA change she presented was so fascinating I thought other boardmembers might appreciate it.

The Kartvelian (or South Caucasian) language family includes Georgian, Mingrelian, Laz, and Svan. Georgian is well-known for having a very strange MSA, which you could probably call mixed-active (it's often cited as an example of a mixed-ergative language, but this isn't really correct), and its sisters too have very unique case-marking patterns. A characteristic trait of this family is that verbal morphology is highly intertwined with case-marking. TMA paradigms ("screeves" in Kartvelological literature) are grouped together into Series that share morphological and syntactic properties. This means that certain screeves call for one pattern of cases and others, another. The details vary by language (and dialect, but I'll just describe the patterns of the standard languages and one particularly unique dialect). Ignore the case nomenclature; the names of the cases (NARrative, NOMinative, and DATive) often don't correspond to their traditional usage—and this is to be expected for a language family with very unusual MSA. I have glossed verbs minimally; keep in mind they generally encode screeve, transitivity, polypersonal agreement, etc.


Svan verbs in Series I (which includes the present and imperfective past screeves) call for an accusative case alignment, but one which doesn't distinguish direct and indirect objects. Note that (2) is an active/unergative intrasitive while (3) is inactive/unaccusative.

(1) giorgi al ambäw-s xaṭūli tamara-s
G.NOM this news-DAT tells T-DAT
"Giorgi tells this news to Tamara"

(2) dede ič̣ḳuārda
mother.NOM was.thinking
"The mother was thinking"

(3) staman ḳwešni
pitcher.NOM breaks
"The pitcher breaks"

In Series II (which includes the aorist or perfective past), however, nouns are marked according to an active alignment. This means transitive and active intransitive subjects are marked one way together and oppose transitive DOs and inactive intransitive subjects.

(4) sosruq̇w-d laxṭix šīra nart-äl-s
S.-NAR returned millstone.NOM Nart-PL-DAT
"Sosruq̇w [mythological hero] returned the millstone to the Narts [mythological giants]"

(5) dede-d adč̣ḳūre
mother-NAR thought
"Mother thought"

(6) šed-är ädḳušurǟnx
dish-PL.NOM broke
"The dishes broke"

Svan MSA can be summarized thusly. (A=transitive subject, Sa=active intrans. subject, So=inact. intrans. subj., O=direct object, G=indirect object)

Code: Select all

           A   Sa  So  O   G


Svan shares this alignment with Old and Modern Georgian.


Standard Laz exhibits active alignment in both Series I and II.

(7) baba-k mečaps skiri-s cxeni
father-NAR gives child-DAT horse.NOM
"The father gives a horse to his child"

(8) bere-k imgars
child-NAR cries
"The child cries"

(9) ḳoči ğurun
man.NOM dies
"The man dies"

(10) baba-k cxeni meču skiri-s
father-NAR horse.NOM gave child-DAT
"The father gave a horse to his child"

(11) bere-k isteru
child-NAR played
"The child played"

(12) ḳoči doğuru
man.NOM died
"The man died"

To summarize:

Code: Select all

           A   Sa  So  O   G


In the Artašen dialect of Laz, at least the following sound changes took place:
s → y / V__#
Cs → y / __#
y → Ø / i __
k → Ø / i__#
Thus, whereas standard Laz has ḳoči, ḳočik, ḳočis "man.NOM, man.NAR, man.DAT", Artašen has ḳoči, ḳoči, ḳoči. In other words, syntactic cases have completely merged, and, at least morphologically, it displays a neutral alignment.


Whereas Laz extended active alignment to all Series, Mingrelian alignment is always accusative. However, there are two concurrent accusative patterns, one that it shares with Svan and Georgian in Series I and a second that evolved from the active alignment its sisters exhibit.

(13) muma arjens cxen-s skua-s
father.NOM gives horse-DAT child-DAT
"The father gives a horse to his child"

(14) jğab-i mušens
girl-NOM works
"The girl works"

(15) ḳoč-i ğuru
man-NOM dies
"The man dies"

(16) muma-k cxen-i meču skua-s
father-NOM horse-NOM gave child-DAT
"The father gave a horse to his child"

(17) jğabi-k sxaṗu
girl-NAR danced
"The girl danced"

(18) ḳoč-k doğuru
man-NAR died
"The man died"


Code: Select all

           A   Sa  So  O   G



In addition to the above case-marking patterns, all Kartvelian languages exhibit inversion, a syntactic process that demotes initial subjects to indirectobjecthood in the dative case. An IO is demoted to obliquehood, usually taking the allative case or a postposition. Initial DOs are promoted to subjecthood according to the principle of unaccusativity. Specifics vary from language to language, but in general experiential verbs (such as love, want, see) and all verbs in Series III (which includes perfect and/or evidential screeves) govern inversion. For example, in Svan:

(19) dis čǟž loxoda gezl-äšd
mother.DAT horse.NOM gave child-ALL
"The mother [apparently] gave her child a horse"

We know the logical subject is the surface IO because it is coindexed on the verb with IO markers; likewise the logical DO triggers subject markers. From Modern Georgian:

(20) m-c̣er-s
"He writes (it) to me"

(21) m-iq̇var-s
"I love him"

(22) Ø-s-c̣er [←*Ø-h-c̣er]
"You write (it) to him"

(23) Ø-u-q̇var-xar [←*Ø-h-iq̇var-xar]
"He loves you"

The morphological implications of this syntactic process are very complicated. For instance, though active intransitives undergo inversion, inactive ones do not. Compare (22) and (23), from Laz, to their Series I and II counterparts above.

(22) bere-s umgarinun
child-DAT has.cried
"The child has cried"

(23) ḳoči doğureleren
man.NOM has.died
"The man [apparently] has died"

I think you could analyze these data such that inversion affects all verbs and calls for another active alignment pattern parallel to that found in Svan, Georgian, and Laz. Harris, however, maintains that it is possible to view inversion synchronically as a ruled-governed syntactic process rather than an independent MSA.

Diachronic Account

Harris gives a very convincing account of the genesis of these coexisting alignment patterns. She posits that in (or up until some stage of) Proto-Kartvelian, case marking was entirely ergative: the ancestor of the modern narrative case was a true ergative case; the modern nominative a true absolutive; and the modern dative a traditional IO-marking dative. The language's "screeve" inventory was probably fairly poor (at least in comparison to its daughter languages) and distinguished aspect most prominently. The punctual aspect (ancestor to modern Series II paradigms) was unmarked, while the durative (Series I) was indicated by a morpheme that may be related to a nominal pluralizer (compare Georgian aḳetebs "he is doing it" and bavšvebi "children").

At some time during Proto-Kartvelian, the durative aspect became associated with a syntactic operation called Object Demotion wherein DOs were demoted to IOs. This had the following consequences: initial ergative arguments became final absolutives (as Object Demotion makes a verb intransitive); initial DOs become marked with the IO-marking case (viz, the dative); and initial IOs become final obliques (though still marked with the dative, or perhaps a case phonologically similar to the dative. Note that the marking of demoted IOs in the dative is common cross-linguistically). Eventually, Object Demotion ceased to function as a productive syntactic rule. The verb is reanalyzed as being unchanged in its transitivity, and the case pattern is reinterpreted as a the nominative-accusative alignment found today in Georgian, Svan, and Mingrelian in Series I.

It is necessary that PK was ergative in the period when Object Demotion was still productive. If it were active, the pattern NAR / DAT / DAT would occur alongside NOM / DAT / DAT in Series I. There are still traces of ergativity to be found in the Kartvelian languages which supports this observation. For example, verbs of motion and position (going, standing up, sitting down) are intransitive and often semantically active, but in most languages/dialects, their subjects are always in the nominative, even in Series which govern active alignment. It is widely observed cross-linguistically the most common words are those which are most likely to retain archaic features (eg, case distinction among English pronouns), so it would make sense that these verbs hold on to a trace of ergativity from the family's past.

However, tracing the shift from an ergative MSA to an active one in Series II is somewhat more difficult than tracing the ergative to accusative shift in Series I. Harris posits that several concurrent factors played a roll in this change, and shifted the alignment gradually. Reinterpretation of transitivity in incorporated object constructions and of unemphatic pronoun drop led to a transition from "generalized" to "restricted" ergativity. Most ergative languages are generalized—subjects of derived intransitives are marked with the absolutive; some, however, are restricted—such subjects are ergative, as the subjects the derive from are. Compare (24)-(25), from the generalized ergative language Yidiɲ, and (25)-(26), from Walbiri, a restricted ergative language.

(24) waguɖa-ŋgu buɲa gibaːl
man-ERG woman.ABS scratched
"The man scratched the woman"

(25) waguːɖa gabaːɖiɲu buɲaː-nda
man.ABS scratched woman-DAT
"The man scratched at the woman"

(26) njuntulu-ḷu npatju pantu-ṇu ŋatju
"You speared me"

(27) njuntuluḷu npatjuḷa pantuṇu ŋatju-ku
2S.ERG AGR speared 1S-DAT
"You speared at me ~ you tried to spear me"

After PK shifted from generalized to restricted ergativity, transitivity-adjusting operations like incorporation eventually became unproductive. The fossilized case marking became reanalyzed and extended for all active intransitives, resulting in the active alignment characteristic of Series II in Georgian, Svan, and Laz.

It is clear that Georgian and Svan inherited this resulting split MSA and did not alter it. In Laz, the active marking of Series II extended into Series I; in Mingrelian, the function of the narrative case in Series II was extended, triggering a shift from active to accusative. Because inversion operates basically the same in all five attested Kartvelian languages, and because it is a such a common syntactic operation (found not only in neighboring languages but also ones like Kannada and Choctaw), Harris reconstructs PK with inversion and does not attempt to trace its development.

In summary, the following morphosyntactic changes are found in the Kartvelian languages or dialects.

Ergative → Accusative (PK Series I)
Ergative → Active (PK Series II)
Accusative → Active (Laz Series I)
Active → Accusative (Mingrelian Series II)
Active → Neutral (Artašen Laz)
Last edited by ná'oolkiłí on Tue Sep 20, 2011 5:35 pm, edited 1 time in total.

User avatar
Posts: 1630
Joined: Wed Mar 08, 2006 5:00 pm
Location: Braunschweig, Germany

Re: Morphosyntactic alignment

Post by WeepingElf »

Wow, an excellent diachronic account of morphosyntactic alignment in Kartvelian, one of the coolest language families on Earth. Rock'n'roll! A million thanks!
...brought to you by the Weeping Elf
Tha cvastam émi cvastam santham amal phelsa. -- Friedrich Schiller
ESTAR-3SG:P human-OBJ only human-OBJ true-OBJ REL-LOC play-3SG:A

Acid Badger
Posts: 196
Joined: Tue May 11, 2010 5:50 pm
Location: Berlin, Germany

Re: Morphosyntactic alignment

Post by Acid Badger »

WeepingElf wrote:Wow, an excellent diachronic account of morphosyntactic alignment in Kartvelian, one of the coolest language families on Earth. Rock'n'roll! A million thanks!
I second that. Thank you very much for sharing it!

Posts: 402
Joined: Tue Jul 25, 2006 10:12 pm

Re: Morphosyntactic alignment

Post by cromulant »

Those Kartvelian MSAs are like intricately-woven tapestries.


Gac is direct-inverse and fluid-S. The ditransitive alignment depends on the hierarchical relationship between the two objects: if R > T, then it's dechticaetiative; if R < T, then double-object.

Here's an idea for an active lang that's been bouncing around in my brain lately:


Nominative-accusative (aka "nominative")
Accusative-dative (aka "dative")


Post Reply