Knowledge predication : A semantic typology

Sammanfattning: The present thesis is an investigation of the cross-linguistic expression of knowledge predication (‘He knows that it’s raining’, ‘she knows the boy’ etc.) Knowledge predication is investigated in parallel texts—specifically New Testament translations—in a genealogically and areally stratified variety sample of 83 languages. A selection of 100 verses, representing the variety of knowledge as well as near-lying concepts, are manually analysed and annotated for lexical and constructional information and analysed using both qualitative and quantitative methods. In the first part of the thesis (chapters 2-4), the theoretical and methodological foundations of the investigation are presented. The definition of knowledge is discussed and a survey of previous research on knowledge and its aspects is presented, drawing on work in philosophy and psychology as well as linguistics. The results are the basis for the selection of verses for investigation. The novel sampling procedure employed—a development of the Diversity Values sampling method with added areal stratification—is introduced. This part of the thesis also situates the investigation within the field of semantic typology, describes the theoretical frameworks used – constructional typology and cognitive semantics—and discusses the methodological challenges of parallel corpora drawing on results from translation studies.In the second part of the thesis (chapters 5-6), the question of whether knowledge is generally distinguished cross-linguistically is addressed. A similarity measure based on lexical and constructional annotations and incorporating automatically inferred paradigmatic relationships is presented and applied. This is then the basis for agglomerative hierarchical clustering, the results of which are interpreted as evidence of an extended knowledge domain (‘know’, ‘understand’ and ‘find out’) and a narrow knowledge domain (‘know’). Also addressed is the question of the universality of knowledge predication, and the sample’s two possible exceptions to this (Kalam [Nuclear Trans New Guinea, PNG] and Fasu [Isolate, PNG]) are discussed in some depth.In the third part of the thesis (chapters 7-11), the cross-linguistic expression of the extended knowledge domain is investigated in more detail. Using Classical Multi-Dimensional Scaling (also known as Torgerson Scaling or Principal Co-Ordinate Analysis), the dimensions along which the expression of knowledge varies cross-linguistically are explored. Their semantics is investigated, and the typology of their expression is discussed. Additionally, the connections of the different parts of the knowledge domain to other domains such as perception—are explored and analysed from a cognitive semantic perspective. Finally, quantitative evidence for knowledge having a core (‘know that’) and a periphery (‘know person’, ‘understand’ etc.) is presented, including a measure of directionality of formal motivational relationships.The fourth part (chapter 12) concludes and summarises the thesis. Six generalisations regarding the linguistic expression of the knowledge domain are made, describing for example restrictions on co-expression patterns. Finally, a semantic map of the domain of knowledge predication is presented, synthesising the results of the thesis, and establishing a basis for future research.