Three continuum traffic models and the simple input-output one are implemented at a signalized intersection and tested against limited data collected by time lapse photography. A comparative performance evaluation is made based on estimatios of the queue length, the effective queue size and interrupted total travel time. These estimates are obtained from initial and boundary conditions measured in short time increments at each end of the link under consideration. The test results indicate that the continuum models are more effective in estimating the above state variables and measure of effectiveness. A simple statistical adjustment suggests that the least sophisticated continuum model performs consistently better than any other alternative. The findings of this study can only be tentative due to the limitations of the data base.