when we are interested in the joint distribution of two random variables, it is useful to have a summary of how much the two random variables depend on each other. The covariance and correlation are attempts to measure that dependence, but they only capture a particular type of dependence, namely linear dependence.