The method I use is to have the object in a Vertex Buffer. Then ProcessVertices into another Vertex Buffer. The second vertex buffer now contains the screen coordinates of each vertex, then I can compare the screen position of the click to the screen position of the vertices to find out which one was clicked. It sounds slow, but it''s really not bad. You do, however, have to make sure when creating your vertex buffers that they are in system memory, not video because reading the transformed vertices from video mem is SLLLLOOOOOOWWWWWW!