Seediq
For a PDF version of my Kuala Lumpur (Oct 2002) handout on Seediq manner-adverbial verbs, click here!
For a PDF version of my Tucson (Feb 2003) handout on Seediq final particles, click here!
For a PDF version of my AFLA 11, Berlin (April 2004) handout on Seediq adverbs and the intraposition problem, click here!
Seediq is an Atayalic language spoken in the mountains of northern Taiwan. It is spoken in two maximally differentiated principal dialects, the Paran dialect centred around Wushe in Nantou County (although most speakers can be found today in the villages of Chingliu and Chungyuan to the north-west of Puli), and the Taroko dialect primarily spoken in the hinterland of Hualien on the Pacific coast.
There are also intermediate dialects which share features with both of the extremes, such as the Toda and Truku dialects spoken in the valleys north-east of Wushe. These share phonological features with the Taroko dialect but various lexical and syntactic features with the Paran dialect. It is quite conceivable that the Toda and Truku dialects as they appear today represent a more original stage of Seediq, while the Paran and Taroko dialects have evolved in diverging directions (in particular as far as case-marking is concerned). Thus, while Toda and Truku have object case-marking on pronouns, Paran has lost this, and Taroko has generalized it to all nouns denoting humans.
A view of a Seediq village
The Seediq people originally practised slash-and-burn agriculture, but the advent of the Japanese brought with it the development of wet-field agriculture. The wet-field is known in Seediq as yqeyaq.
Typological facts about Seediq
Word order
Seediq has the typologically unusual word order VOS. This is shared primarily with its closest relatives in Taiwan, as well as a substantial group of other Austronesian languages, comprising, among others, Malagasy and Fijian.
Voice
The Western Austronesian voice system, known in Austronesianist circles as the focus system, differs from voice in a language such as English in various ways. Firstly, instead of having one "passive", as English has, Seediq has three "passives".
In English there are two voice forms, Active, which indicates that the Actor is subject, and Passive, which indicates that the Patient is subject (both of these definitions are rather inexact, but will serve for our purpose here). In Seediq, on the other hand, there are four voice forms, which indicate that the clause subject is Actor, Patient, Location or Instrument, respectively. (Again, these semantic roles don't really paint the correct picture, but will do as an approximation - Holmer 1999a presents a much more detailed and exact description of what the various voice forms actually "do".)
An even more important characteristic of voice in a language like Seediq is that the "passives" are not "passive" in one crucial sense, i.e. they do not involve detransitivization (that is, the deletion of the Agent). Thus, while the "passives" crucially do indicate that the clause subject (the NP in nominative case) is not the Actor, they are not functionally equivalent to "passives" in a language like English. In all three Seediq "passives", the Agent is obligatorily present, and may not be deleted. Nor is it marginalized in a prepositional phrase like an English by-phrase, it is and remains a core argument of the verb, and is realized in what is most plausibly identified with ergative case (actually identical to genitive in Seediq and other languages of the same type).
These facts have caused various influential researchers (notably Starosta) to simply label languages such as these "ergative languages". There are facts which support such an analysis, and facts which do not, and certainly these languages do have ergative characteristics, although they differ in important respects from other types of ergative languages. Given that the morphological variation of the verb corresponds to a change in the surface argument structure of the verb, I consider the terms "diathesis" or "voice" suitable in this context, but note that my view of voice differs greatly from that of other GB syntacticians (if you don't believe me, see my thesis!).
Clitic pronouns vs. agreement
I commented on the question of Atayalic bound pronouns in Holmer 1993, and suggested that a treatment as clitics solves word order problems. Since then, Chang Yung-Li has suggested in his 1997 thesis (and other work) that these bound pronouns should be seen as agreement affixes, since they can cooccur with a clause-mate full NP (which is normally not the case with clitics). Chang is right, of course, that cooccurring with clausemate full NP's (which Seediq bound pronouns clearly do) is not typical behaviour of clitics. At the same time, these bound pronouns are not affixal in their phonological behaviour (they do not cause changes in the stress pattern of Seediq verbs, as other clitics do). Therefore, as I suggest in my review of his thesis, they should be anaysed as agreement morphemes syntactically speaking and clitics morphologically speaking. This is another point where Seediq syntax proves to be a fascinating topic to work with.
Hopefully, this young boy will grow up speaking Seediq
