OBJECTIVE: Investigate the relationship between on-field markers of concussion severity and postinjury neuropsychological and symptom presentation in an athlete-specific population. DESIGN: Case control study. SETTING: Multicenter analysis of high school and college athletes. PARTICIPANTS: A total of 78 athletes sustaining sports-related concussion were selected from a larger sample of 139 concussed athletes. ASSESSMENT OF PREDICTOR VARIABLES: On-field presence of disorientation, posttraumatic amnesia, retrograde amnesia, and loss of consciousness. MAIN OUTCOME MEASURES: ImPACT, a computerized neuropsychological test battery, was administered pre-season and, on average, 2 days postinjury. Good postinjury presentation (n = 44) was defined as no measurable change, relative to baseline, in terms of both ImPACT memory and symptom composite scores. Poor presentation (n = 34) was defined as a 10-point increase in symptom reporting and 10-point decrease in memory functioning (exceeding the 80% confidence interval for measurement error on ImPACT). Athletes failing to meet good or poor selection criteria (n = 61) were not included in the analysis. RESULTS: Odds ratios revealed that athletes demonstrating poor presentation at 2 days postinjury were over 10 times more likely (P < 0.001) to have exhibited retrograde amnesia following concussive injury when compared with athletes exhibiting good presentation. Similarly, athletes with poor presentation were over 4 times more likely (P < 0.013) to have exhibited posttraumatic amnesia and at least 5 minutes of mental status change. There were no differences between good and poor presentation groups in terms of on-field loss of consciousness. CONCLUSIONS: The presence of amnesia, not loss of consciousness, appears predictive of symptom and neurocognitive deficits following concussion in athletes. Athletes presenting with on-field amnesia should undergo comprehensive and individualized assessment prior to returning to sport participation. Continued refinement of sports concussion grading scales is warranted in lieu of consistent findings that brief loss of consciousness is not predictive of concussion injury severity.