A stochastic approach has been presented for superplastic deformation of Ti-6Al-4V alloy, and probability functions are used to model the heterogeneous phase distributions. The experimentally observed spatial correlation functions are developed, and microstructural evolutions together with superplastic deformation behavior have been investigated by means of the two-point and three-point probability functions. The results have shown that the probability varies approximately linearly with separation distance, and deformation enhanced probability changes during the process. The stress-strain behavior with the evolutions of probability function can be correctly predicted by the model. The finite element implementation using Monte Carlo simulation associated with reconstructed microstructures shows that better agreement with experimental data of failure strain on the test specimen.