Theoretical considerations and simulation studies have led to the expectation that patient dose in CT (computed tomography) can be reduced significantly without a concomitant loss in image quality if tube current is modulated according to rotation angle-dependent x-ray attenuation. In this study, the simulation results presented in Part I were validated with phantoms. We used one cylindrical, two oval, and one elliptical phantom, available both as mathematical descriptions and in physical form, to mimic different parts of the human anatomy. Prototype hardware was available to control tube current on a commercial clinical CT scanner. The potential for dose reduction was evaluated for sinusoidal and attenuation-based current modulation for variable modulation amplitudes. Agreement between simulations and measured results was better than within 10%. Dose reduction values of 8%-56% were found depending on the phantom geometry and tube current modulation function. Attenuation-based tube current modulation consistently yielded higher reduction than fixed-shape sinusoidal modulation functions. For the shoulder phantom and 70% modulation amplitude, 44.6% dose reduction was measured as compared to 34.1% for sinusoidal modulation. A maximum of 56% was measured for the shoulder phantom including inserts. Specifying mAs reduction as an estimate for dose reduction proved to be a valid and conservative estimate; measured dose is reduced more strongly than the total mAs product both centrally and on average. First patient studies confirm the results of simulation and phantom studies. We conclude that attenuation-based online tube current control has great potential for reducing patient dose in CT and that it should be made generally available for clinical use.