Abstract
It is shown that by a proper selection of the probabilities of 0 and 1 at the inputs, the efficiency of random test generation can be improved. This correspondence includes some results describing the testing of actual logic networks used in a computer.