Abstract
A new procedure is proposed for finding a transformation that produces a random variable whose distribution is symmetrical. Studies show that the procedure is much more efficient than the best currently available, when the symmetrical distribution has either very heavy or light tails.