Wireless sensor network is one of the effective tools for data gathering in a variety of environments which has a lot of challenges. Controlling the congestion is one of the challenges that primarily depends on the way the large number of sensors in a network are organized and controlled. A group of nodes is called a cluster which sends information to the gateway through the cluster head. This structure helps the sensor networks to operate smoothly (i.e., without any congestion) under light or normal load and then a significant part of the network suddenly becomes active in response to the event of interest. During the event of interest the sensor nodes generate a huge amount of message which causes congestion in the network. The traditional solution of this problem is to decrease the source generation rate. Usually in cluster network the congestion happens in the cluster head which forwards all messages of the cluster and downstream clusters' messages. A cluster head can reduce some message load by shifting some its sensors to neighbouring cluster. A lightly loaded neighbor cluster head expands the cluster by receiving some sensors to reduce the load of the congested neighbor cluster. In this way, if a number of cluster heads on the downstream routing path of the congested cluster head apply the same process then the total amount of message reduction creates a significant effect towards reducing the congestion. This paper introduces the cluster shrinking and expending technique in a special way considering the congestion in the network. Simulation results are presented to verify the effectiveness of the proposed cluster shrinking and expanding technique.