Three studies on childhood accidents are presented. The aim was to study the importance of different factors regarding the accidents in question. The following factors have then been taken into consideration: the enviromental hazard, the accident proneness, the supervision and the education. Methodologically the investigations were carried out with an epidemiological technique. One is of a descriptive nature and the other two more analytically oriented. The studies are based on two different 1-year-materials consisting of accidents among children recorded in the emergency departments of Ostersund Hospital and the University Hospital in Uppsala. The results indicate that risk factors in the children's physical milieu played an important role in the occurrence of the accidents: clearly identifiable risk factors in the environment could be connected with 52% of the accidents, whereas some deficiency in supervision was noted in 20%. The investigators could identify a number of specific risk factors. Attempts were made to examine how frequency and type of childhood accidents vary with the population structure and social structure in well-defined housing areas, but the results were hard to evaluate because of methodological problems. The results are presented against the background of a detailed discussion on central methodological problems in epidemiological accident research. It is pointed out in particular that epidemiological methods have clear limitations in attempts at studying the low-frequency events that each individual type of accident in fact comprises. It is of great importance that in future research, side by side with the traditional epidemiological methods, other techniques are tested with the aim of obtaining maximal usable information from a detailed study of individual accidents and their backgrounds.