A novel three-dimensional (3D) self-localization scheme APIT-3D (approximate point-in-te-trahedron) was presented for wireless sensor networks (WSNs). The target sensor performed APIT-3D test to judge whether it resided inside or outside of tetrahedrons, which narrowed down the possible location areas (PLA). Then, APIT-3D averaged the centers of gravity (COG) of the intersection of PLA as the unknown sensor-s estimated location. Simulation results indicate that, as a range-free scheme, APIT-3D can achieve location accuracy within 40% of radio range, while the communication cost remains reasonable comparing with its 2D counterpart. Since APIT-3D depends on no ranging measures or global infrastructure, it presents a low-cost solution for self-localization of densely deployed and large scale WSNs in 3D spaces.