The analysis of correlations between financial assets is an important basis of modern finance.Either the selection of portfolio or risk management involves the estimating and forecast of correlations,so the quest for reliable estimates of correlations has been the hot issue of concern for many researchers.This paper examines the correlations between China's A-shares and B-shares markets using the DCC-MGARCH model proposed by Engle in 2002.The empirical results show that: the correlations are overall positive and obviously time-varying;the dynamic correlations are relatively low and more volatile from 1997 to 2001,but they obviously increase and have less volatility after 2001,which indicates that the integration of A-shares and B-shares in Chinese stock market is strengthening;even though,the correlations have not rised obviously after the high position;as a whole,the correlations between A-shares and B-shares are relatively low and the nature of market segmentation is still evident.Finally,we employ behavioral finance theory to analyze and explain the dynamic correlations between A-shares and B-shares and its time-varying character.