Neurosurgical registrars are frequently called upon by A&E staff and physicians to interpret emergency head CT (computed tomography) scans out of hours. This appears to reflect the reduced threshold for scanning patients and the nonavailability of a radiologist to report these scans. This study was undertaken to assess the safety of such practices. Five neurosurgical registrars, blinded to each other and to the radiology reports, interpreted 50 consecutive emergency head CT scans (both trauma and nontrauma) from the hospital's imaging system as a pilot study. These were initially graded as normal or abnormal. Abnormal scans were assessed for the presence of an intracranial bleed, pneumocephalus, skull fractures, cerebral contusions, mass effect, midline shift, ischaemia or hydrocephalus. The agreement of the observers' recordings with the report issued or approved by a consultant radiologist was evaluated using SPSS Version 13.0. Four of the five registrars assessed a further 150 scans in a similar manner to complete the study. There was a good general agreement between the formal reports and the neurosurgical registrars' identification of normal scans (average Kappa 0.79). The radiology reports and the registrars also agreed well on the presence or absence of intracranial blood, contusions and pneumocephalus (Kappa value > 0.70). The agreement was poorer for ischaemia, mass lesions (other than intracranial haematomas), grey white differentiation, evidence of raised intracranial pressure and midline shift (Kappa < 0.5). Neurosurgical registrars compared well with radiologists when it came to assessing emergency head CT scans as normal or detecting a surgical lesion. The agreement was poorer on subtle abnormalities. The practice of neurosurgical registrars informally 'reporting' on emergency head CT scans cannot be recommended.