Headway or inter-arrival time of vehicles is an important parameter in traffic flow, especially in urban roads, since this is one of the main parameters to determine the minimum (safe) gap between vehicles and the capacity. The studies done on headway analysis have mostly concentrated on homogeneous traffic where the flow follows lane discipline. This ideal situation does not exist on Indian urban roads, where the traffic is very heterogeneous, and do not follow lane discipline. A study of headway at different but similar locations of urban four lane divided (two lanes for each direction of flow) roads of a metropolitan city in India (Chennai) was made. The peak hour flow at these locations were observed to be high (varying from 3189 to 9987 vehicles). The study of headway, after removing the 5 % of long headways, at these location indicate that Log normal 2, Inverse Gauss and the Exponential distributions are the most appropriate ones for these conditions of flow. An attempt was made to study the individual inter arrival time of each category of vehicles in the traffic stream. Extensive data extraction was done and the analysis of headway of categorized traffic was made. Appropriate distribution for each category is recommended. The study will be a good input for modeling vehicle generation in simulation studies.