Abstract: Gridlock is an extreme traffic state where vehicle cannot move at all. This research studies the development of gridlock by theoretical and numerical analysis. It is shown that the development of gridlock can be divided into several stages. The core of the development is the evolution of congestion loop. A congestion loop is comprised of a number of consecutively connected spillover links. The evolution of a congestion loop always tends to be stable, i.e. the state of all related links tends to be identical.. Under the stable condition, traffic states of all links are identical. A novel concept, “virtual signal” is proposed to describe the queue propagation and spillover during the stabilization. Simulation results show that congestion propagates in an accelerated way. The prevention of the first congestion loop is crucial. The achieved results have potential use for future network traffic control design and field applications