Correlations between international equity markets are often claimed to increase during periods of high volatility. Therefore the benefits of international diversification are reduced when they are most needed, i.e. during turbulent periods. This paper investigates the relationship between international correlation and stock-market turbulence. We estimate a multivariate Markov-switching model, in which the correlation matrix varies across regimes. Subsequently, we test the null hypothesis that correlations are regime-independent. Using weekly stock returns for the S&P, the DAX and the FTSE over the period 1988?99, we find that international correlations significantly increased during turbulent periods.