The purpose of this work is to investigate the method to judge a vehicle whether it is free flowing one or not, and to apply the result to vehicular headway distribution models, particularly, to formulate headway distribution parameters as a function of traffic flow level. A number of authors have developed two component headway distribution models, and the basic assumption of models is vehicles on the road can be devided into two groups according whether or not they are interacting with the vehicle ahead. This assumption is agreeable, but practically the fraction of free or constrained vehicles is estimated without clear definition of free or constrained vehicle. Headway distribution models give a good fit to the data which used to estimate model parameters, but not assure estimated model parameters are applicable to other data under comparable conditions, and if functional formulas of models are changed the fraction of free or constrained vehicles may change. In this paper, α(t), the fraction free flowing vehicles which follow the vehicle ahead with time headway t, are defined by use of variance and covariance of relative speed distribution, and headway distribution is divided into free and constrained part. For both of free and constrained part, translated longnormal distribution are used, and the new models by use of either free or constrained headway distribution model are developed. All headway distribution model parameters are formulated as a function of traffic volume. It becomes clear that traffic flowing patterns under free flow condition and congested flow condition are fundamentally different. All data used in this paper are collected for no-passing zone of rural two lane roads.