Final considerations Investigative recommendations are discussed here. This research was a combination of literature study and accident study in order to check literature results against the actual Dutch practice. This provides several useful pointers for infrastructural, traffic engineering and transportation policies: guide cyclists through residential areas instead of along arterial roads; restrict the number of two-way bike paths and use compensatory measures in case these are used; use speed reductors in order to slow down vehicles entering and leaving side-roads. These measures make it possible to realise a substantial reduction in the number of accidents. It is advisable to actively promote these measures among road maintenance authorities. In addition the study provides a number of less unequivocal recommendations. In some cases relations were found which could however not be completely substantiated statistically. In other recommendations there were reasons based on actual practice to refrain from making unequivocal recommendations. The recommendations may be used in a possible revision of Ontwerpwijzer Fietsverkeer. In case of a number of these recommendations a follow-up study might be advisable. The use of centre guides, for instance, appears to contribute to the safety at arterial roads with more than two lanes, where the centre guide provides waiting room for cyclists. An observational study might provide more information. Subsequent studies may also be useful for recommendations about markings. This study also provides two important recommendations about accident studies concerning the safety of cyclists: I. always take into account the volumes of cyclists and motorists in studies concerning the relation between infrastructural characteristics and cycling accidents. The volumes of cyclists and motorists are major predictors of accidents at intersections: the more cyclists pass an intersection, the more cyclists may be involved in an accident; the more motorists pass an intersection, the more possible collision partners. Although this appears to be self-evident, accident studies very often do not compensate for volume. When a group of intersections with characteristic A is compared to another group with characteristic B there will almost always be small or large differences in volume between both groups, both in cyclists and in motorists. If no correction is made for volume, differences in numbers of accidents per intersection due to volume differences may wrongly be attributed to infrastructural characteristics. II. Only conduct a study into the relation between infrastructural characteristics and cycling accidents at a national level, to guarantee a minimum study size. This study investigated over 500 intersections in 7 towns. Despite this size the number of accidents was barely enough to allow reliable pronouncements. Evaluating a specific measure, for instance a particular type of speed reductor, is impossible at a local level. Consultants might provide the specialist expertise required for an evaluation, but the number of accidents at a local level is too small. The second point is related to the previous paragraph. Deviations in particular circumstances can in most cases be substantiated by local and regional customisation or an integral assessment in combination with other interests. A structural deviation from parts of measures can usually not be substantiated to a sufficient degree by road maintenance authorities as their research opportunities are too limited for the purpose. Studies at a local level may best focus on possible improvements in particular situations or areas, for instance with the aid of the AVOC method (Approach Traffic Accidents Concentration points) or a qualitative research method, a number of which have been described in the Toolkit analysemethoden; Toepassing kwalitatieve analysemethoden verkeersveiligheid (Van den Bosch, 2008)