Capacity analysis of signalized intersections basically consists of estimating saturation flow and delay. Pre-timed signals are most commonly used in developing countries. This research deals with development of saturation flow and delay models for pre-timed signalized intersections with reference to non-lane based traffic condition prevailing in Bangladesh. In order to account non-uniformity in the static and dynamic characteristics of the vehicles passenger car unit (PCU) values for each vehicle is found out using synchronous regression technique and a range of site-specific PCU values were obtained. From this study, it has been observed that unified PCU concept does not hold good for non-lane based traffic condition and it has been recommended that the analysis should be site specific for non-lane based traffic condition. The saturation flow for each study approach was calculated using the average PCU values and multiple linear regression techniques were then used to derive predictive saturation flow models. Field delay for each approach is calculated based on HCM 2000 guidelines. It has been observed that HCM 2000 delay model consistently over estimate delay at degree of saturation more than 1.0. It has been suggested from the analysis that theoretical incremental delay (due to random arrival and over saturated queues) in HCM 2000 delay model be reduced by 70 % to better reflect field conditions in capacity analysis for non lane based traffic condition.