BACKGROUND: Historical papers on the treatment of congenital dislocation of the hip suggest the use of preliminary traction to facilitate closed reduction or to decrease the risk of avascular necrosis (AVN) of the femoral head. In the 1980s, some authors questioned the role of preliminary traction and suspended its use, yielding satisfactory results. Since then, several studies called into question this method, and some authors have continued to recommend preliminary traction while other authors have discouraged its use. MATERIALS AND METHODS: We reanalysed the full set of radiographs of 71 hips (52 patients) surgically treated by a medial approach after 4 weeks of preoperative longitudinal traction. The mean age at operation was 16 months. Before and after traction, the height of the dislocation was graded according to the Gage and Winter method. The hips were divided into two groups: group 1, in which the traction was effective, and group 2, in which the traction was not effective. These two groups were statistically analysed regarding the severity of the dislocation, the age of the patient at surgery and the incidence of AVN. RESULTS: Preliminary traction was effective in 48 hips (68%, group 1), while it was not effective in the remaining 23 (32%, group 2). The effectiveness of preliminary traction was statistically related to the height of the dislocation and to the age of the patient at surgery, with traction being less effective in more severe dislocations and in older children. The incidence of AVN was statistically lower in group 1 than in group 2. CONCLUSIONS: In our study population, despite not having a control group, preliminary traction-when effective-seemed to reduce the incidence of AVN in patients surgically treated for congenital dislocation of the hip. The effectiveness of the traction was influenced by the severity of the dislocation and the age of the patient; it worked better for less severe dislocations and in younger children. To reduce hospital costs, traction should be applied at home. LEVEL OF EVIDENCE: 3.