This research focuses on analysis of the grammatical functions between an adnoun clause and a noun phrase in Korean. The key task is to determine the relation between two constituents in terms of one of the various functional categories such as subject, object, complement, and appositive. The problem is mainly caused by the fact that functional morpheme, crucial for identifying the relation, is not used in the noun phrase. We propose a statistical method in determining the grammatical functions. A tagged corpus is used for training a new model and a smoothing technique known as "backed off" is employed to tackle the data sparse problem. As experimental tests prove our model to be useful, we make it assist error detection of parsing results.