A method for generating random-dot textures is described that provides statistical control of any n adjacent points (n-grams), while leaving constant the statistics of shorter spans. The method thus allows the experimenter to isolate statistics of n-grams of any span length to determine the nature of their influence on texture discrimination. Variables that control phase are relatively unimportant. The most significant variables are constraints imposed upon span lengths less than 3 that regulate gray level and spatial frequency content. However, span lengths of 3 or greater may still influence discrimination by altering the distribution of the spatial-frequency content.