Three models proposed for the growth mechanism of hydrogenated microcrystalline silicon films (mu c-Si:H) from reactive (silane and hydrogen mixture) plasmas are reviewed. The 'etching model' is discussed using experimentally obtained relationship between radical generation rate in plasmas and growth rate of films. The 'chemical annealing model' is investigated through the growth of films using a layer-by-layer method with and without cathode shutter. Substrate-temperature dependence of crystallinity of the resulting films and initial growth behavior of silicon films on atomically flat GaAs substrate clearly support the 'surface diffusion model'.