A review of published mathematical models used to simulate air sparging is provided. Applicability of the models, efforts to test the models using experimental data and contributions of modeling efforts to the practice of air sparging are also discussed. Compartmentalized lumped-parameter models and multiphase flow models have dominated air-sparging modeling efforts. In essence, each class of models requires the assumption of a continuum over some model domain. Each approach has significant benefits as well as some inherent disadvantages. Based on the literature, both lumped-parameter modeling and multiphase-flow modeling have been successful in improving our theoretical understanding of the air-sparging process and in facilitating practical development of sparging systems. Lumped-parameter models are simpler to use, and can lend considerable insight to sparging operations. Multiphase flow models have the potential to offer a more realistic simulation of the airflow process, but may require a considerable amount of data collection for model input. The literature suggests that for any air-sparging model to be useful for field applications, detailed model calibration is necessary. It is recommended that models incorporate, in some fashion, the diffusion and dispersion of contaminants to macro-scale air channels, and nonequilibrium interphase mass transfer of contaminants. These mass-transfer-limited processes are frequently listed as causes for the "tailing" of vapor-extraction effluent contaminant concentrations that are frequently observed during field applications. However, time-varying mixing of relatively clean and contaminated vapors in the extraction system may also explain this tailing. Geophysical imaging techniques and inverse modeling combined with air-sparging pilot tests and measurement of traditional hydrogeologic parameters may allow for successful modeling efforts.