汶水方言是泰雅語群中最為保守的方言，它有許多獨特的現象(包括音韻、構詞、句法）不見於其他泰雅群的方言。這些現象中有一些無疑反映了古泰雅語的面貌，從而使汶水方言在泰雅語群中的歷史研究乃至於泰雅和其他南島語相互關係的探索上，都佔據了一個最關鍵的地位。

本文討論汶水方言的下列具體特徵：(一)句中每一個名詞都必須使用格標記；(二)區別普通名詞和人名；(三)區別特指名詞和非特指名詞；(四)主語總是出現在句末；(五)人稱代詞附著式中，其主格和屬格各分兩套(區分的條件可能是句法的或音韻的，都有所討論）；(六)有許多助詞不見於其他泰雅方言，(七)主格為有定，受格為無定。以上這些特徵，(三)、(五)、(六)為汶水方言所獨有，似為後起的現象，而(一)、(二)、(七)則為一般西部南島語所習見卻為其他泰雅方言所遺落(尤其(一)、(二)），似為存古的現象。過去許多學者由於缺乏像汶水方言的佐證，便以為泰雅語的發展與一般西部南島語大異其趣，甚而主張應於古南島語之下別立泰雅語群一支。這種看法恐怕已經不能成立。作者曾從音韻、詞彙方面，舉汶水方言為證，加以反駁。本文更增加了句法方面的證據，使這項反駁更形完整。

In this paper I describe all the case markers and the various pronominal forms that mark cases and grammatical relations in Mayrinax, the most conservative dialect of Atayal. These case markers indicate the following different functions: nominative, accusative, genitive, benefactive, dative, instrumental and locative. I shall also compare the differences in the case markers as used in Mayrinax and Squliq, a better known and more innovative dialect of Atayal.

The Mayrinax dialect of Atayal is conservative not only in phonology and morphology, but also in syntax. It retains all the case-marking particles that have generally been lost in the other dialects of Atayal, as well as in Sediq. It has an obligatory case marker for each noun in every sentence. It distinguishes between a common noun and a personal noun. Since such phenomena are wide spread in western Austronesian languages and in some Formosan languages, Atayal at an earlier stage should resemble the other languages more closely. In other words, Atayal has not been as aberrant a language as is generally believed.

Mayrinax also manifests many interesting grammatical features not found in the other dialects of Atayal or in any other Formosan language. It has a complex case-marking system. In addition to the distinction between common and personal nouns, it also distinguishes between specific and non-specific common nouns. Moreover, it has a much more elaborate pronominal system than any other Atayal dialect has. It has two different sets of short forms for both the nominative and genitive pronouns. The word order problem, especially for pronouns, is intriguing.

Some of the syntactic features in Mayrinax must have been inherited from the parent language, while the others may have been innovations of its own.

This paper also provides some solid evidence to show that the nominative is definite, whereas the accusative is indefinite in Mayrinax. Although such a distinction is well known among western Austronesian languages, evidence is still unavailable for most other Formosan languages.