By computing the residual cross-correlation function (CCF) ， this paper attempts to explore the nonlinear causality and spillover effect of volatility of the series of the official exchange rates of China，Europe and Japan，respectively，against the USA dollar by employing daily data from Jul． 21，2005 to Dec． 23，2011．Test statistics of nonlinear causality of conditional variance developed by Yin-Wong Cheung，Lilian K． Ng and Hong Y． are used to examine the nonlinear causal relationship between the variances of standardization residual of the three kinds of exchange rates． There upon，a BEKK-MGAＲCH model is built． Finally，the nonlinear causality and spillover effect of volatility on these exchange rates in post financial crisis era are analyzed and interpreted，and the robustness of the nonlinear causal relationship between the variances of time series is verified． he embedding of the truncated kernel function in Hong test causes the low order lag terms a given greater weight，thereby accurately portrays the characteristics that recent fluctuations have a larger impact on current fluctuations． Prior to the establishment of the Vector GAＲCH model，few literatures tested the nonlinear causal relationship first，even though some conducted a simple test，they mistakenly treated Q2(p) as the test statistic of the nonlinear causal relationship between the variances of variable sequences．