The image quality and dose savings of self-masking subtraction tomosynthesis (SST), which combines digital tomosynthesis with subtraction of a blurred self-mask, was tested. High-quality images of the inner ear of a head phantom were obtained at moderate dose savings. Although they were taken with linear motion, they did not exhibit the streaking due to off-fulcrum objects that is characteristic of conventional linear tomography. SST could reduce patient dose by a factor of at least 12 in examinations of the inner ear, and the mechanical aspects can be implemented with moderate modifications of existing instrumentation.