This paper reports on the real-data testing of a real-time adaptive freeway traffic state estimator that is based on macroscopic traffic flow modelling and extended Kalman filtering. The testing intends to demonstrate some main features of the estimator that are partly due to its adaptive capability based on on-line model parameter estimation. These features are (1) avoiding off-line model calibration; (2) adaptive to changing environmental conditions; (3) enabling incident alarms. The reported testing results are quite satisfactory and promising for future applications of the estimator.