The mechanism of two-phase flow through fractures is of importance in understanding many geologic processes. Currently, two-phase flow through fractures is still poorly understood. In this study, nitrogen-water experiments were done on both smooth and rough parallel plates to determine the governing flow mechanism for fractures and the appropriate methodology for data analysis. The experiments were done using a glass plate to allow visualization of flow. Digital video recording allowed instantaneous measurement of pressure, flow rate and saturation. Saturation was computed using image analysis techniques. The experiments showed that gas and liquid phases flow through fractures in nonuniform separate channels. The localized channels change with time as each phase path undergoes continues breaking and reforming due to invasion of the other phase. The stability of the phase paths is dependent on liquid and gas flow rate ratio. This mechanism holds true for over a range of saturation for both smooth and rough fractures. In imbibition for rough-walled fractures, another mechanism similar to wave-like flow in pipes was also observed. The data from the experiments were analyzed using Darcy's law and using the concept of friction factor and equivalent Reynold's number for two-phase flow. For both smooth- and rough-walled fractures a clear relationship between relative permeability and saturation was seen. The calculated relative permeability curves follow Corey-type behavior and can be modeled using Honarpour expressions. The sum of the relative permeabilities is not equal one, indicating phase interference. The equivalent homogeneous single-phase approach did not give satisfactory representation of flow through fractures. The graphs of experimentally derived friction factor with the modified Reynolds number do not reveal a distinctive linear relationship.