Perhaps if we hold the position where the spine attaches to the dendrite constant and just vary (by some parameter) how far away the spine can grow. This way we are just changing the growth direction of the spines, and not where they are located on the dendrites.