The traffic loading of long-span bridges is governed by congestion. Real-world observations show that congestion can take several different forms. Nevertheless, most previous studies on bridge traffic loading consider only queues of vehicles at minimum bumper-to-bumper distances. In fact, such full-stop queues are rare events, while in most cases congestion waves propagate through the traffic stream, so that on a bridge there are periodically times of closely-spaced vehicle concentrations and times of flowing traffic, where vehicles are more distant. In this paper, an acknowledged traffic micro-simulation model is used for generating congested traffic on a single-lane roadway encompassing two bridges (200 and 1000 m long). Two truck percentages are considered (20% and 50%) and different congestion patterns are analysed in relation to their traffic features and effects on bridge loading. It is found that for the case of 200 m span and 20% trucks slow-moving traffic results in greater loading than full-stop conditions. Finally, the frequency of occurrence of different forms of congestion is taken into account based on recent available data, rather than being assumed as in most previous research. It is found that considering only the widely-used full-stop conditions leads to an over-estimation of the characteristic total load by about 10% for the cases of 200 m span with 50% trucks, and 1000 m with 20% trucks; for the case of 1000 m span with 50% trucks, the over-estimation drops to nearly 5%. However, for the case of 200 m span with 20% trucks, considering only the full-stop conditions leads to a slight under-estimation of the total load.