Impulse response functions are useful for studying the interactions between variables in a vector autoregressive model. They represent the reactions of the variables to shocks hitting the system. It is often not clear, however, which shocks are relevant for studying specific economic problems. Therefore structural information has to be used to specify meaningful shocks. Structural vector autoregressive models and the estimation of impulse responses are discussed and extensions to models with cointegrated variables or nonlinear features are considered.