One major limitation of existing microscopic traffic simulation models is the lack of data about vehicle interactions during the lane-changing process in congested merging areas. The interactions between vehicles during lane changing in the congested weaving sections are quantified. A series of acceleration-deceleration models was developed for the merging vehicle, its putative leader (PL), and its putative follower (PF). After US-101 data were analyzed, the yielding behavior of the merging vehicle's PL for merging cooperation and the lateral separation between vehicles in the weaving section (features largely ignored in previous research) were introduced into the proposed acceleration—deceleration models, and visual angle data were used as the stimuli to reflect the influence of longitudinal and lateral vehicle movements. Car-following behavior was incorporated into the model (e.g., the PF gradually changed its car-following leader from the PL to the merging vehicle, dependent on relative vehicle locations). Calibrated and validated with FHWA's next generation simulation (NGSIM) program trajectory data, the proposed acceleration-deceleration models could qualitatively simulate the behavior of drivers in lane-changing maneuvers with acceptable training and testing errors. To verify the cooperative behavior of the PL in lane changing, the proposed PL acceleration-deceleration model was compared with a base car-following model. Results indicate that incorporating the effects of the merging vehicle on the PL can enhance the realism of a lane-changing model. Findings from this study can contribute to the understanding of interactions between vehicles during complex lane-changing processes.