This paper evaluates two different versions of Zwicker [Acustica 6, 365–381 (1956)] and Maiwald’s [Acustica 18, 193–207 (1967)] excitation‐pattern model for intensity discrimination. The single‐band version is that originally used by Zwicker and Maiwald. They assumed that performance was determined by the critical band in which the excitation grows most rapidly with increasing level of the stimulus. The multiband version forms an optimum decision based on information in all critical bands. Predictions from the two versions of the model are compared to data for intensity discrimination of tones as a function of level and frequency, for partially masked tones and for white noise. In general, the single‐band version yields predictions in qualitative agreement, but not in quantitative agreement with the intensity‐resolution data for pulsed tones. The multiband version yields predictions in good qualitative as well as quantitative agreement with the data, except at high frequencies.