This paper discusses an automatic way to derive a type hierarchy for verbal items in Korean based on their subcategorization. There are three steps: First, all the dependent categories of the each verb are extracted from the Sejong Treebank. Second, based on the frequency of the dependent categories of each verb, the most stable subcategorization frames are selected, and two statistical measures are tested with some variations in their cutoff values. The resulting subcategorization frames are then compared with those from the Sejong Electronic dictionary for evaluation. The final step is to form a type hierarchy for Korean verbal items, based on the chosen subcategorization information.