The effective use of computer-generated pictures as a trial-unique probe for studying the visual memory is described. The shape of the pattern is determined by means of a fractal algorithm with pseudorandom parameters. This method enables us to easily obtain thousands of moderately complex and sufficiently diversified pictures in series from a given number which serves as the seed of a pseudorandom number generator. We can thereby create a new and unique set of pictures if a new seed is given, as well as retrieve exactly the same pictures in the same sequence as when the original seed is given. These properties eliminate the demand for the massive memory space in a computer otherwise needed to store the entire set of stimulus pictures.