Several studies indicate that road users are not able to distinguish roads according to official road categories. As a means of designing a sustainably safe traffic environment the concept of Self-Explaining-Roads (SER) has been developed. In order to support safe driving behaviour and appropriate speed choice drivers should be enabled to recognize the type of road they are on, so they know the adequate behaviour. However, most Dutch roads appear to display not enough structure to allow for an effective utilization of these SER-principles yet. The present study investigated a method that can be used to find dimensions (characteristics) that contribute to better recognizable and distinguishable road categories. Five experiments were performed to investigate how subjects distinguished and recognized different environments categories. A sixth experiment was conducted to link these environment characteristics with subjects' estimates of their driving speed. Experiment I was a pre-test to select equally salient levels from the different dimensions, so that categorization in subsequent experiments could not be based on salience differences within dimensions. In subsequent experiments subjects were to judge whether each of a large number of environments was a member of an (artificial) class. They were not informed on what dimensions they had to do this. They gradually learned how to classify based upon feedback after each trial. To concentrate on basic classification principles first, Experiment 2 and 3 used artificial environments as stimuli. Experiment 2 tested how difficult it was for subjects to learn to classify an environments on the basis of differences within a single dimension only. Experiment 3 tested to what extent subjects are able to correctly classify environments when one out of three relevant dimensions provided deviant information. Results showed that subjects were able to learn categorizing on the basis of one or two dimensions, but could not correctly learn to use information from three stimulus dimensions together. They used two out of three available dimensions, and ignored the third one. Experiment 4 and 5 were similar to Experiments 2 and 3, but now with road environments as stimuli. Results strongly resembled the Experiments 2 and 3 results. Subjects only used the two stimulus dimensions that were learnt most easily, and ignored the third. Subjects' estimated own driving speeds in the road environments presented (Experiment 6) were, on average, about 92 km/h. Neither Centre-line marking (interrupted/continuous), Road surface (light/dark grey), or Reflector posts (present/absent) influenced these estimated driving speeds. Both Lane width (2 vs. 3.6 m) and Red bicycle lanes (present/absent) considerably affected estimated driving speeds with the narrow lane width and presence of bicycle lanes having the lowest estimated speeds. The results have implications for road design. In practice, road categories are defined by a set of characteristics on several dimensions such as lane width, road surface, delineation, and elements along the road. Due to local circumstances some of the pre-defined characteristics can not always be applied in a given traffic situation. The present results suggest that providing redundant information does not help the individual road users since they only use a few of the available characteristics for individual road classification. Correct classification by road users can only be enabled if the few dimensions that are used by road users always provide consistent and correct information on the type of road. It appears to be best to identify those dimensions such that differences among categories are learnt most easily. Only specific dimensions seem to affect drivers' speed choice directly. (A)
Abstract