Kavalan, an Austronesian language spoken in Taiwan, has been variously analyzed as accusative, ergative, and split ergative. These different conclusions stem from the fact that certain two-argument clause patterns are ambiguous regarding transitivity. To settle the matter, it is necessary to distinguish canonical transitive clauses from dyadic intransitive clauses. In this paper, we evaluate three proposals that have been made concerning Kavalan transitivity and actancy structure in terms of their morphosyntactic and semantic properties. We pay special attention to the form tu and determine that it is best analyzed as an oblique marker rather than as an accusative marker. We also conclude that there is only one canonical transitive construction, that found in two-argument -an clauses. The two-argument m- clauses, commonly analyzed as canonical transitives in most previous analyses, are treated as extended intransitives or pseudo-transitives--a type of intransitive clause. This leads to the conclusion that Kav alan is best analyzed as a purely ergative language. 1. INTRODUCTION (1) In the studies of Formosan, Philippine, and other western Austronesian languages as well, (2) the distinction between valency and transitivity has often been neglected. Many Austronesianists equate monadic clauses with intransitive clauses, and dyadic clauses with transitive clauses without considering the relevant morphosyntactic and semantic properties that each exhibits. As a result, many such languages are analyzed as having two distinct types of transitive constructions and an unconditional split-ergative system, something that is typologically unusual. Kavalan, an Austronesian language spoken in Taiwan, has been commonly analyzed in this way. This paper reviews these analyses from a broad typological perspective and determines the canonical transitive construction and actancy structure of Kavalan based on morphosyntactic and semantic criteria. We hope thereby to shed new light on the study of western Austronesian syntax generally. We begin with a brief introduction to our theoretical orientation in section 2. We summarize previous analyses of Kavalan transitivity and actancy structure in section 3. Finally, in section 4, we evaluate three proposals concerning Kavalan transitivity and actancy based on morphosyntactic and semantic criteria (see Hopper and Thompson [1980] and Gibson and Starosta [1990] for detailed discussion). It is shown that the form tu is best analyzed as an oblique marker rather than as an accusative marker. Moreover, based on the morphosyntactic and semantic properties that Kavalan clauses exhibit, we conclude that there is only one canonical transitive construction in Kavalan, that involving two-argument -an clauses. Two argument m- clauses, which have been commonly analyzed as canonical transitives in previous analyses, are treated as extended intransitives or pseudo-transitives, a type of intransitive clause. By analyzing the dyadic -an clauses as canonical transitives and the dyadic m- clauses as extended intra nsitives, we will conclude that Kavalan is best analyzed as a pure ergative language, rather than as an accusative or split-ergative language. 2. THEORETICAL ORIENTATION (3) In this paper, we employ a revised version of Dixon's Basic Linguistic Theory to describe Kavalan clause structure. Some notions that are crucial to the discussion of Kavalan clause structures are discussed in this section. 2.1 CORE ARGUMENTS VS. PERIPHERAL ARGUMENTS Basic Linguistic Theory as outlined in Dixon (1979, 1994) and Dixon and Aikhenvald (2000) distinguishes core arguments from peripheral arguments (also called adjuncts). The occurrence of core arguments is determined by the head (usually a verb) of a clause. The core arguments must be stated (or be understood from the context) for a clause to be acceptable. …