To define a subset of injured children for whom emergency cervical spine radiography may be unnecessary, we performed a retrospective chart and radiologic review. Two entry methods were used: All injured children, from birth through 16 years, who had received cervical spine radiographs at The Children's Memorial Hospital from September 1983, to September 1984, were included. All patients from birth to 16 years with proven or suspected cases of cervical spine injury who had received cervical spine radiographs and who had been treated at either the Children's Memorial Hospital or the Northwestern University Spine Trauma Unit during period 1974 to 1984 also were included. Each child's chart was reviewed, and 84 clinical variables were recorded. All radiographs were reviewed by a pediatric neuroradiologist. Of 206 children studied, 59 had cervical spine injuries. A clinical algorithm was derived using the following eight variables: neck pain; neck tenderness; limitation of neck mobility; history of trauma to the neck; and abnormalities of reflexes, strength, sensation, or mental status. The following decision rule was selected: Positive findings in any of these eight variables mandates cervical spine radiography. This algorithm correctly identified 58 of 59 children with cervical spine injury, yielding a sensitivity of 98% and specificity of 54%. Cervical spine radiographs could have been avoided in 79 children (38% of the entire sample). This algorithm performed better than did models derived from logistic regression analysis of the same data. Validation trials are required prior to the implementation of this or other clinical decision algorithms in practice.