Models are an important tool for decision making. However, in order to get proper results, these models must be validated and only be used in situations where the conditions of the validation apply. Blind trust on a model can lead to unexpected and inaccurate results. Advancements can be made to reduce the number of situations where this occurs. Not only by making the models more accurate, but also by doing more field studies for validation of behavioural aspects of the traffic. One of these aspects is the process of lane changing and car following behaviour. These two aspects determine the general longitudinal and lateral driving behaviour. Mathematical models that describe these types of movements for each individual vehicle provide the building blocks for microscopic simulation. In most models, these two aspects are modelled independently, but newer models, such as the integrated driving behaviour model (Toledo, 2003), attempt to mould this into an integral decision structure. This research attempts to validate the lane changing and car following behaviour of three models: FOSIM, VISSIM and the aforementioned Integrated driving behaviour model. These models are compared against a dataset from TNO of the motorway A270, in situations where free flow conditions apply. The models are tested on the desired speed distribution, the merging point distribution, the accepted gap distributions and the lane change distribution. The lane changes that are being found are classified by their distinctive causes, the so-called “triggers”. Six triggers are defined for lane-change classification. The main result is that calibration and validation play a major role in the validity of the models. For all tested simulation packages, their default parameters did not reflect the observed data. This means that the driver’s attitude and the traffic conditions have a large impact on the general driver behaviour. In free-flow traffic conditions, Dutch drivers tend to be risk-averse, as reflected in the low number of voluntary lane changes and the wide gap acceptance distribution. This risk-averseness is usually not part of a model’s default parameter set and therefore calibration is essential to simulate the traffic correctly. Furthermore, the different triggers helped to get a clearer view about what type of lane changes occur, where, and why they occur. The FOSIM simulation results show that this model has serious limitations. A main point is that this model is too deterministic regarding driver characteristics. Although in theory probabilistic factors could be added to the model, further advancements of the model, such as implementing probabilistic behaviour, requires reprogramming of the simulation package, which was not possible within this research. VISSIM gave better results, but it over-estimates the number of voluntary lane changes in free flow conditions on Dutch motorways when using the default behavioural parameters. Further calibration of these parameters did partially correct this error, but the remaining estimation errors differ per voluntary lane change trigger; courtesy and speed gain related lane changes are under-estimated while lane changes to keep right were over-estimated. Furthermore, the gap acceptance behaviour was not much improved. This may indicate that other boundary conditions, such as traffic generation, where wrongly assumed in the simulation. Also, one could argue if a gap selection algorithm could improve the accuracy. Further research is required to test these hypotheses. The Integrated driver behaviour model could not be completed within the time constraints of this research, but analysis of the car-following aspect of this model shows that this model has some limitations that could be easily solved by several counter-measures. Driver observation and acceleration behaviour issues could be solved by integrating psycho-physiological factors into the model, such as observation thresholds and multiple acceleration regimes. A main recommendation is to perform more validation research of current models to gather more calibrated parameter sets for a wide range of traffic conditions. The used data collection method, road side cameras, was an accurate enough method to gather enough data for this research. This method can be widely applied for other researches too with different camera mounting points, such as lamp posts and sign gantries. The triggers that have been defined in this research could be used in other studies to find if there are differences in driver behaviour for each trigger. However, within this research, 10 to 15 percent of the lane changes could not be classified in one of the six triggers. This may indicate that there is either a classification error or a missing trigger. For VISSIM, ranges of recommended parameter values for Dutch traffic in free-flow have been found and are provided in this study.