Binaural loudness matching and intermodal magnitude matching experiments were performed to test systematically the origins of the phenomenon observed earlier that the variability of binaural loudness matches was larger when sound intensity was varied in the normal ear than when it was varied in the contralateral ear with raised threshold and loudness recruitment. In the experiments, the raised threshold and loudness recruitment were produced by masking a 1‐kHz tone with narrow‐band random noise. In intermodal experiments, magnitude matches were performed between the masked tone and the length of lines projected on a translucent window pan. The results are consistent with the earlier observations. They show in addition that the variability depends on the slope of the matching functions in a complicated, not previously anticipated way, irrespective of whether the functions are intra‐ or intermodal. More specifically, for moderate slopes, the variability in the ear with loudness recruitment decreased as the slope increased. The reverse was true for the unmasked ear or the line length—when the slope was large, the variability increased with the slope. Since the variability decreased in one ear and increased in the other, the ratio of the variabilities increased as the slope increased. When the slope was equal to one, both variabilities tended to be the same.