Pickering emulsions have been developed into many applications. The droplet size is always the most fundamental property influencing the system's stability, rheology and reactivity. This thesis describes the investigation into three aspects of droplet size in Pickering emulsions—the quantification of size dispersion, the understanding of the unusual droplet size distribution caused by multi-body coalescence, and the study of droplet size effects in the oil-water separation using magnetic carbon nanotubes (MCNTs).To quantify size dispersion, researchers have proposed four different uniformity indices. Here, we quantitatively relate the index span, defined as the normalized size range containing 80% of the total particle volume centered around the volumetric median, to the coefficient of variation and then to the other three uniformity indices. We show that a consistent criterion for mono-dispersity can be obtained with span for all types of distribution functions. We further show that span has an explicit mathematical relationship with another index defined as the ratio between the volume-weighted deviation and median, which suggests that the latter index can also be interpreted unequivocally on the statistical basis like span.Next, we discover a magic droplet size distribition with distinctive maxima related to each other through the cubic root of tetrahedral numbers in Pickering emulsions. This size distribution is attributed to an unusual behavior of multi-body coalescence in an ensemble of Pickering droplets, where all closely packed nearest neighbors are involved in a single coalescence event. In addition, we show that the multi-body coalescence is promoted by the repulsive interactions among the particles at interface but inhibited by the attractive interactions.Last, we report the kinetic analysis of treating dodecane-made simulated produced water using MCNTs. This technology first wraps the droplets with MCNTs and then pulls the wrapped droplets out of water in a magnetic field. Here, we find that the wrapping kinetic model follows the first order of both MCNT and oil concentrations. We also show that the salinity improves both the efficiencies and the kinetics of wrapping. The increased performance is quantitatively attributed to the corresponding increasing dodecane droplet size with increasing salinities.