Many different classification schemes have been used in the analysis of road traffic accidents but the agreement between coders using the same classification scheme is rarely tested and/or reported. As a high intercoder agreement is a prerequisite for a study’s validity, this is a serious shortcoming. The aim of the present study was, therefore, to test the intercoder agreement of the Driving Reliability and Error Analysis Method (DREAM) version 3.0 by letting seven coders from different European countries analyse and classify the causes of the same four accident scenarios. The results showed that the intercoder agreement for genotypes (contributing factors) ranges from 74% to 94% with an average of 83%, while for phenotypes (observable effects) it ranges from 57% to 100% with an average of 78%. The results also showed that weaknesses in classification schemes, methods, training of coders as well as in presentation of accident information can be identified by testing the intercoder agreement.