PURPOSES : Traffic congestion on freeway generally occurs when the traffic volume exceeds the road capacity. Most traffic manuals *such as the Korean Highway Capacity Manual) present the highway capacity as approximately 2,000 units/hour. However, in the real world, freeway congestion occurs for various reasons, including unusual driver behaviors, physical road limitations, and large traffic volumes. Thus, the flow rate at a traffic breakdown can have a wide range of volumes. Therefore, using 5-min volume and speed data from the field, this study explores the stochastic features of traffic breakdowns on major urban freeways in Seoul. METHODS : First, a breakdown point is defined by applying a wavelet transform to identify the sharp drop in the speed data near freeway bottlenecks. Second, based on the flow rate at and before a breakpoint, a survival analysis is performed to construct the probability distributions of the traffic breakdown. Log-rank tests are also conducted to verify the similarities of the distributions between freeways. RESULTS : The analysis results confirm the stochastic features of the urban freeways in Seoul. Specifically, the freeways have typical S-shaped distributions of breakdown probabilities. However, the distributions rise steeply (exceeding a 50% of breakdown probability) at flow rates of 1,150 vphpl to 1,700 vphpl; this is lower than the general expectation. CONCLUSIONS : The statistical differences in the probability distributions for freeways indicates that applying a general standard to every urban highway could raise problems. This study has a limitation in identifying the specific causes of traffic congestion owing to the by physical relationships between individual vehicles. An investigation if vehicle trajectory data should be conducted to examine these aspects in further detail.