We present a theoretical approach for understanding the interrelations between dynamics and structure of signal transduction pathways. We consider large sets of networks with a specific number of kinases and phosphatases. Our methods are based on nonlinear differential equations and pathway dynamics is characterised in terms of signal amplification and signal duration. We show that networks with a high number of kinases, high connectivities and low phosphatase activities tend to be unstable and run, therefore, the risk to display autoactivation. Analysis of signal transduction pathways retrieved from databases reveals that several structural characteristics required for pathway stability are fulfilled for networks of very large size.