This study examines the effects of random level variation, a method used in studies of profile analysis [3-6,14,15]. Presentation levels for a complex of sinusoids are varied randomly on each interval of a two-interval, forced-choice detection task in which subjects are required to detect an increment on one of the sinusoidal components of the complex. Three experiments are reported. The first experiment examines the effect of the range of level variation. The second is concerned with the effects of the median level about which the presentation levels vary. The third experiment is designed to provide a within-trial analysis of the effect of the differences in presentation levels. As the range of level variation is increased, ability to detect the increment decreases. The results indicate that detection performance is best at moderate intensity levels and decreases at lower and higher levels. Finally, the difference in levels within a single trial has little if any effect.