本論文提出將韓文的推測表現分為證據推測表現與認知性推測表現之三種測試，即是指i 與現實相反假設測試、ii 根據之存在測試、iii 證據之存在測試。前述i,ii 是將川添等（2010）對於區分日文推測表現時所提案之測試翻譯為韓文，並透過本研究確認此測試亦可用於韓文。此外iii 為本論文全新之提案，期待亦可應用於日文。藉此顯現以上三種測試可運用於語言共通之推測表現分類測試之可能性。此外，本論文試圖將韓文的推測表現，從上述之意思性測試獨立，依其型態特徵分類。首先，將韓文的推測表現分為A類型（含冠形詞形語尾的推測表現）與B類型（不含冠形詞形語尾的推測表現）。而B類型內核心之推測表現（推測表現內含冠形詞形語尾的推測表現以外部份），再分為僅出現冠形詞形語尾ul共存類型與冠形詞形語尾ul un nun ten全可同時共存類型。本論文並指出這些形式之分類與前述意思性分類有密切關係。換言之，關於B類型，除少數例外，含有ul之推測表現為認知性推測表現，含有ul un nun ten之推測表現則為證據推測表現。關於例外，本論文亦指出其成為例外之理由。

In this paper, the following three tests are presented in order to classify modal expressions in Korean into evidential expressions or epistemic expressions: i. counterfactual test, ii. existence-of-reason test, iii. existence-of-evidence test. In Kawazoe et al. (2010), we already proposed that the same tests as i and ii are useful to classify Japanese modal expressions. In this study, we showed that these tests can be applied also to the categorization in Korean. The test ”iii” is first introduced in this paper and may be applied to Japanese. Therefore this study suggests that those three can be useful tests of modal expressions cross-linguistically.The classification of Korean modal expressions by their formal characters was made as well as the above mentioned classification by their meanings. First, we have divided the expressions into Type A, which does not include an adnominalizer, and Type B, which does. Then Type B into the one which includes the core-modal-expressions that can only co-occur with the adnominalizer -ul, and the other which includes those that can co-occur with each adnominalizer: -ul, -un, -nun, or -ten. We have then pointed out that there are close correspondences between these formal classifications and semantic ones, the latter of which were already discussed. We have argued that essentially the modal expressions which include -ul are epistemic and those which include -un, -nun, or -ten are evidential, and have given an account for the exceptional cases.