This paper is concerned with the problem of exploiting pause information for recovering dependency structures of read Japanese sentences. In our past work, two kinds of pauses were employed: post-phrase pause that immediately succeeds a phrase in question, and post-postphrase pause which immediately succeeds the phrase that follows a phrase in question. It was found that simultaneous use of two kinds of pause information improves the parsing accuracy compared to the case where only the post-phrase pause is used. In this paper, we employed yet another kind of pause (pre-phrase pause), which immediately precedes a phrase in question. By combining the three kinds of pause information appropriately, the parsing accuracy was further improved compared to the case where the post-phrase pause and the post-postphrase pause were used as in our previous work.