Well in 3d stuff, all vertices are based on the origin. The x,y,z is an offset in object space from the object's origin. This way, when poser transform/translate the object, the matrix has the object to world transform and the object's offset in the world. They just zip the vertices through this matrix and voila, world space vertices. The can also multiply matrices so, for example, the figure has a figure matrix, this times the neck matrix gives the neck in WS, this times the head matrix gives the head vertices in WS and so forth.