The effect of double reading on the number of breast cancers detected and the number of women recalled was examined in 15 457 women aged 50-59 years attending for mammographic breast screening. All the screening mammograms were interpreted independently by two radiologists without knowledge of each other's findings. All the cases selected by either radiologist were then reviewed by both radiologists and a consensus decision made on whom to recall for further studies. This method of double reading increased the number of breast cancers detected by 9% (P < 0.05) and decreased the number of women recalled by 45% (P < 0.001). Double reading of mammography screening films together with consensus decisions on the selected cases is recommended.