Sonodynamic therapy is a promising modality for cancer treatment based on the synergistic effect of cell killing by a combination of drugs (sonosensitizers) and ultrasound. The effectiveness of sonodynamic therapy was demonstrated in cell studies and in tumor-bearing animals. The mechanism of drug-dependent sonosensitization is unknown, but it seems likely that various mechanisms operate for different classes of sonosensitizers. Ultrasound-mediated sonodynamic activation of porphyrins is particularly poorly understood and hypotheses ranging from singlet oxygen production to peroxyl radical formation have been advanced. Evidence has been obtained against some of the arguments which were proposed in favor of the singlet oxygen hypothesis and the requirement of extracellular localization of gallium porphyrin ATX-70 molecules for sonosensitization in HL-525 cells has been established. Short-lived toxic intermediates produced from sonosensitizers by ultrasound are implicated in the mechanism of sonodynamic activation of some sensitizers. EPR spectroscopy was used to identify proposed intermediates in the activation of certain sonosensitizers (e.g., N,N-dimethylformamide, DMSO) by ultrasound. Certain water-soluble azo compounds, which are stable at hyperthermia temperatures and can be decomposed by ultrasound to give peroxyl radicals capable of initiating peroxidation of lipids and damaging other cellular sites, appear to be promising sonosensitizers.