本論文是要研究中文如何處理名詞論元之訊息流動，以瞭解中文訊息結構
和其文法之間的複雜關係。名詞論元分為三大類：（１) A (及物動詞之
主事者），（２）O （及物動詞之受事者），（３）S（不及物動詞之唯
一論元）。S 與 O 又可各細分成「動詞前」的 S或 O、「動詞後」的 S
或 O 兩種。訊息結構是針對以下的屬性作研究：（１）「新／舊訊息」
，（２）「可辨認性」，（３）「泛指／特指」，（４）「人／非人」，
（５）「主題連續性」 。為了探討各類言談篇章的訊息結構是否存有個
別差異，本論文採用了三種口語語料，包括一篇約四十分鐘的日常對話，
以及兩種故事的敘述。首先，由名詞帶出的新／舊訊息所形成的訊息結構
反映了中文的詞序：絕大部分的舊訊息出現在動詞前之論元位置，無論
是 A、 S 或O；新訊息則大多置於動詞後，不論是 S 或 O的位置。三類
口語語料都獲得一致的結果。可是，以往的類似研究顯示其他語言都以論
元為中心，組成了 ergative (S/O 與 A 相對） 或 accusative (A/ S
與 O 相對）型態的訊息結構，都不受制於詞序。本論文利用中文語料，
發現訊息結構並非單一的系統。不同的語言不但有著不一樣的結構型態，
它跟語法的關係更是密切。以中文為例，它的語法跟訊息結構一樣，都不
受制於論元本身。除了「新／舊訊息」的屬性，三類口語語料對於其他的
訊息性質所形成的結構都沒有一致的結果。它們的功能在於顯示各類言談
篇章間存有之個別差異。另外，「人／非人」這語意屬性的文法分布跟訊
息流動和語法更是沒有關係，因為它不像新／舊訊息般受制於詞序；語法
也不是針對這個屬性而運作。跟它有密切關係的是「主題連續性」。帶有
「人」屬性的名詞大多是說話的主題，出現的論元位置也多是動詞前的
A 或 S 這兩種。總之，論元在中文不是形成訊息結構、主題連續性、語
法運作的首要因素。中文的言談篇章與語法之間呈現對應 （iconicity）
的關係。

In this study, I investigate how Mandarin Chinese speakers
manage the information flow of nominal referents across
grammatical roles in different discourse, in trying to
understand the intricate relationship between Chinese discourse
and the shape of its grammar. The conversation text and the
two types of oral narratives are taken as distinct types of
speaking for purposes of examining the potential text- type
difference against a cluster of information flow properties:
information status, identifiability, generality, humanness and
topic continuity. The S and O roles in the preverbal position
are also distinguished from those in the postverbal position.
The unique and recurrent patterning of information status
across conversation and narrative discourse suggests a
pragmatic motivation for word order. The preverbal roles, be
it A, S or O, have a great propensity for given information;
the postverbal roles, either S or O, maintain a much higher
incidence of new information. This lack of accusative or
ergative alignment of grammatical roles has parallel in syntax
where the major syntactic processes treat A, S and O in the
same way. The inconsistent patterning of identifiability,
generality and humanness across the three texts reveals that
they are measures of text- type difference. Semantics in terms
of humanness is rather related to another aspect of discourse--
topic continuity. Not only are humans most likely to be
mentioned again as topics, but human nominals in whatever
grammatical role also tend to re-appear in A or S where humans
typically take place. Non-humans, on the other hand, are
usually brought up again in O where referents of this type
readily occur. To sum up, the fact that grammatical role is not
the primary determinant of information patterning, topicality
and syntactic processes indicates an iconicity between
discourse and syntax.
In this study, I investigate how Mandarin Chinese speakers