Urban travel times are intrinsically uncertain due to a lot of stochastic characteristics of traffic, especially at signalized intersections. A single travel time does not have much meaning and is not informative to drivers or traffic managers. The range of travel times is large such that certain travel times can occur for congested conditions as well as off-peak situations. Therefore, it is better to consider the whole distribution of travel times. This thesis started with the investigation of different monitoring techniques for measuring urban travel times. The challenges in these technologies (e.g., ANPR, Bluetooth devices and probe vehicles) were discussed. From the empirical data, we can observe that urban travel times are rather variable. However, the investigation of travel time variability as done by most researchers is just in a phenomenological descriptive way by fitting some distribution functions (e.g., log-normal, gamma) to observed travel times. The problem arises when applying these distributions to different traffic conditions since they are only calibrated for a specific traffic situation. Therefore, this thesis focuses on developing a theoretical travel time distribution model which can explain fundamental mechanisms of urban travel times. In this thesis, an analytical link travel time distribution model, which considers stochastic traffic processes at intersections, was developed. The model was further extended to a trip travel time distribution model with two intersections, taking into account of traffic signal coordination between these intersections. The parameters in the model were calibrated with VISSIM simulation data, and the validation of the model was done both with VISSIM simulation data and real life GPS data.