The objective of experiments reported here was to analyze the effect of feature interactions on discrimination performance with complex, noise-like sounds. Sixteen pairs of laboratory-generated sounds were used. Acoustic features composing the stimuli included octave bands of noise and amplitude modulation of noise bands by a 10-Hz square wave. Within each sound pair, signals differed by one or more dichotomous features, features present in one signal and not in the other. The thesis presents a feature extraction model for discrimination, incorporating concepts from signal detection theory and auditory information processing. Interactions between features are hypothesized wherein the presence of one feature affects the detectability of another.