On Assigning Semantic Cases in Finnish 1 Urpo Nikanne University of Helsinki 1. INTRODUCTION The Finnish case system is traditionally taken to consist of two types of cases: (i) structural cases: nominative, genitive, partitive, and accusative; (ii) semantic cases, which are divided into two subgroups: (a) locative cases: (aa) internal locative cases: inessive, elative, illative; (ab) external locative cases: adessive, ablative, allative; (ac) general locative cases: essive and translative; (b) marginal cases: abessive, instructive, comitative, prolative etc. (See the introduction of this volume.) It is often assumed that the structural cases reflect the syntactic relations and semantic cases the thematic relations of the sentence. I will argue that all case marking is structural. Following Jackendoff (1987a) I assume that language processing is a computational activity. In language interpretation one translates the phonological structures into a syntactic format, and then the syntactic representation into a conceptual format. The supposed Logical Structure of Language Processing (Jackendoff 1987a: 92) is phonological structure <-> syntactic structure <-> conceptual structure. 2 The function of case marking is, then, to make the relevant level of syntactic representation recoverable. 2. THE CATEGORY OF THE "SEMANTIC CASE PHRASE" IS PP The category of the phrase with semantic case inflection has been a problem for Finnish grammarians. It has the distribution of a PP while there seems to be no Ρ around. For instance: (1) a. Elina käveli kohti koti+a Elina walked toward home+PAR 'Elina walked toward home' This article is based on section 3.2 of Nikanne (1991), which, in turn, is based on the syntactic half of Nikanne (1989). I am grateful to Anders Holmberg, Ray Jackendoff, Joan Maling, Hannu Reime, Trond Trostemd, and Anne Vainikka among others for being interested in my work and for many enlightning discussions. Still, my friends are not responsible for my mistakes, faults and limbos. 2 Of course, the sentence does not need to be analyzed in its entirety to the next level before the process can continue. Jackendoff (1987a: 91-120) discusses the problems of processing in detail. Brought to you by | provisional account Unauthenticated Download Date | 10/6/16 3:49 PM