In this paper we present an algorithm for the (semi-)automatic identification of anaphors whose antecedents are verbal phrases, clauses or discourse segments in Danish Dialogues. Although these anaphors are quite frequent, especially in conversations, they are usually been neglected in computational linguistics. The algorithm we propose contains defeasible rules for distinguishing these anaphors from those who have individual nominals as antecedents. The rules have been identified by looking at the occurrences of these types of anaphor in the transcriptions of two dialogue collections. The algorithm has been manually tested on four Danish dialogues and the obtained results have been evaluated.