RATIONALE AND OBJECTIVES. The authors assessed whether a non-sleep-deprived, second-year diagnostic radiology resident assigned to an after-hours “night stalker” emergency radiology (ER) rotation in an urban university hospital has a measurable impact on the number and clinical significance of “missed” radiologic findings. METHODS. After-hours Emergency Department (ED) radiographs interpreted by radiology residents between January and December 1991 were reviewed daily by ER faculty. Faculty-modified final interpretations were recorded on a worksheet and given to the attending ED physician (EDP). The EDP reviewed and, if indicated, modified clinical dispositions, and categorized missed diagnoses as requiring recall into the following categories: 1 = immediately, 2 = in 24 to 48 hours, 3 = no recall necessary, or 4 = recognized during patient visit by clinicians. Morbidity attributable to “misses” was graded A to C (A, definite; B, possible; C, none). All cases requiring patient recall were evaluated monthly with follow-up information and classified as false-positive, false-negative, or indeterminant. The relative performance of control (traditional “call”) and night stalker groups were compared. RESULTS. Of 26,421 on-call examinations in 1991, there were 489 (1.1%) misses, of which 202 (1.4%) were from the June-to-December study group. The control group residents averaged 2.75 hours of sleep per call night. On night stalker days, on-call residents and the night stalker averaged 5.75 and 7.25 hours of sleep daily, respectively. The fractions (and number) of recall assignments of discordant cases for the control and night stalker groups, respectively, were: 1) immediate 48% (23) and 26% (32); 2) within 48 hours 31% (15) and 26% (31); 3) no recall 79% (38) and 36% (43); and, 4) abnormality not missed by EDPs 10% (5) and 12% (15). Morbidity for the control and night stalker groups, respectively, were: 1) 4% and 0%; 2) 31% and 30%; and 3) 65% and 70%. The amount of rework between July and December 1991 spent by the EDPs to re-evaluate cases because of discordant opinions was more than 68 hours, with no significant difference noted between the study groups. Errors were false-negative, 84.9% (415); false-positive, 7% (34); and indeterminant, 8.2% (40). Radiology faculty errors contributed 5.8% (13) of patient recalls (false-positive, 11; false-negative, 2). Finally, 58/78 questionnaire respondents believed that service quality had improved. No one believed that the standard of service had been lowered. CONCLUSIONS. A dedicated night-shift ER coverage of a busy urban ED improves quality, appropriateness, and timeliness of patient care.