The distribution of droplet sizes produced in a microfluidic T-junction is modeled using the gamma probability distribution. The proposed model is validated using physical insight from the hypothesized mechanism of droplet breakup and statistically using the Anderson-Darling test. Where signal-to-noise ratio is low and the gamma distribution does not hold, an alternative probabilistic description is used to estimate the true mean of droplet sizes. In addition, a new correlation is developed to relate mean droplet sizes to T-junction inlet flow rates and dispersed-phase viscosity. (C) 2016 American Institute of Chemical Engineers