Vehicular time headway is of fundamental importance in traffic engineering. For any traffic simulation to effectively address traffic problems, accurate vehicle generation is essential. Previous researches on this subject have focused solely on the stochastic modeling of uninterrupted facilities. Yet little research has been conducted on interrupted facilities such as suburban arterials. This paper proposes theoretical headway models for a suburban arterial. Using a traffic detector based on laser sensors, a large amount of accurate headway data were obtained in Korea. To analyze the data according to different states of traffic flow, the headways were categorized into five flow groups. Subsequent runs tests for each headway group rejected the randomness assumption for the lowest flow (5-9 veh/min). Therefore, theoretical modeling was performed only for the four remaining flows (10-14, 15-19, 20-24, and 25-29 veh/min). The Johnson SB model achieved the best fit for a flow of 10-14 veh/min, whereas the Johnson SU model provided the best fits for the three other flows. The Log-Logistic and Log-normal models were also accepted for high flows. In addition to analyses on statistics of the collected headway, the characteristics of the fitted model revealed that the headway has somewhat different features, compared to those of uninterrupted facilities. This paper provides a better understanding of headway for interrupted facilities and constitutes a starting point for developing a descriptively accurate simulation model for signalized arterials.