Self-organization appears as a new paradigm for providing efficient, self-adaptive, scalable, fault-tolerant and robust communication protocols in dynamic and distributed wire- less multi-hop networks such as Wireless Sensor Networks (WSNs). Thus, it allows high level applications or proto- cols to operate efficiently above an underlying logical topol- ogy without neither waste of resources nor centralized point of control. A variety of distributed algorithms for form- ing virtual network topologies are being considered as ba- sic mechanisms to achieve these goals. This paper presents performance evaluation and analysis work of two kinds of self-organization algorithms for data gathering applications under realistic radio assumptions for WSN using network simulation tool. In this paper, we investigate the charac- teristics of self-organization algorithms such as connected dominating set and link pruning strategies during the WSN life. We study the latency, the dissipated energy during the chaotic network deployment, the robustness and the cardi- nality of virtual topologies. Then, we study the throughput and the communication overhead of these self-organization algorithms while data gathering application are proposed with one or several sinks.