One of the most important characteristics linked with sound quality and acoustic comfort is the overall perceived loudness. Existing loudness models are applicable only to diotic sounds in spite of the fact that human beings hear dichotic sounds in usual situations. Although binaural recording tools have been developed rapidly and real dichotic signals can be obtained, there are still few dichotic loudness models for the calculation of overall perceived loudness from binaural signals. Traditionally, the mean of loudness values calculated from left and right ear signals has been approximately used for the overall perceived loudness. In this study, the dependence of overall loudness perception on the interaural level differences was investigated by the subjective tests. In the test, subjects should find the diotic signals, of which equally loud dichotic signals are perceivable. In the first test, pure tones with critical band rates of 1, 2.5, 8.5, 14.5, 20.5, 22, and 23 Bark were tested. The larger the interaural level difference, the louder the perception than the mean of calculated loudnesses from left and right ears and the lower the critical band rate, the louder the perception than the mean value. The perception is louder than the mean in the cases that interaural loudness ratios are larger than about 1.2 and critical band rates are lower than about 20.5 Bark. In the second test, five pure tones of 1 kHz having different reference levels of 40, 55, 70, 85, and 100 dB were tested. The larger the interaural level difference, the louder the audible perception than the mean and the lower critical band rate and the louder the perception than the mean value. The perception is louder than the mean in the cases that interaural loudness ratios are larger than about 1.2 and reference levels are lower than about 85 dB. The loudness model was modified appropriately, that the model was applicable to dichotic sounds by using the equivalent diotic levels. The characteristic of the modified loudness model was investigated by the subjective test for the pink noises. It is observed that the loudness values calculated from the present modified model agreed well with the test results. The present dichotic loudness model was applied to several practical dichotic sounds. It is observed that the modified dichotic loudness values are meaningfully different from the mean values in the cases that interaural level differences are relatively large, i.e., hair dryer noises.