本篇論文從認知語意學的角度來探討分類詞的本質，並且結合了Fillmore (1975) 的框架 (frame) 理論和Langacker (1987) 的顯影(profiling) 概念，作為分析的理論基礎，說明語言如何透過分類詞機制來凸顯名詞在認知上的某些語意特徵。本論文所嘗試提出的理論架構最終目的在於解釋所有的分類詞語言，但為提昇語料的可信度，語料來源以台灣華語中的分類詞為主。 研究分析主要建構於名物框架 (N-frame) 的認知基礎上，進而檢視台灣華語中不同類別之名詞藉由不同之分類詞所彰顯的語意特徵 (feature)，同樣重要的是不同的語意特徵在台灣華語分類詞系統中的優先排序。最後我們從其他的分類詞語言中，如閩南語及日語，論證名物框架系及彰顯機制統的普遍性；唯不同的分類詞語言或不同方言中所選擇顯影的語意特徵以及其優先排序都可能有所不同，分類詞系統也因此呈現複雜多樣的面貌。 / On the basis of the theory of Frame, proposed by Fillmore (1975), and the models of profile and base segregation, proposed by Langcker (1987), this thesis aims at a cognitive approach to explain how languages manifest certain cognitive features of a noun by means of classifiers. In addition, the theoretical framework tries to explain all the known classifier languages and to foster a greater accountability and reliability of the data, this study focuses on Taiwan Mandarin. The data analysis is based on the cognitive foundation of an N-frame, whereby it is found that Taiwan Mandarin categorizes nouns by profiling different features of nouns through the use of different classifiers. A hierarchy or ranking of classifier features in Taiwan Mandarin is also provided. Finally, in all known classifier languages, such as Southern Min and Japanese, it is found that the application of the use of an N-frame is consistent and universal in human cognition. The only thing that varies is the profiling system that each language employs and the hierarchy of features it selects. Thus, the use of classifiers manifests varied and complex aspects.