BACKGROUND: Evidence-based clinical data are required for safe return to play after concussion in sport. PURPOSE: The objective of this study was to describe the natural history of concussion in sport and identify clinical features associated with more severe concussive injury, using return-to-sport decisions as a surrogate measure of injury severity. STUDY DESIGN: Cohort study (prognosis); Level of evidence, 3. METHODS: Male elite senior, elite junior, and community-based Australian Rules football players had preseason baseline cognitive testing (Digit Symbol Substitution Test, Trail-Making Test-Part B, and CogSport computerized test battery). Players were recruited into the study after a concussive injury sustained while playing football. Concussed players were tested serially until all clinical features of their injury had resolved. RESULTS: Of 1015 players, 88 concussions were observed in 78 players. Concussion-associated symptoms lasted an average of 48.6 hours (95% confidence interval, 39.5-57.7 hours) with delayed return to sport correlated with > or = 4 symptoms, headache lasting > or = 60 hours, or self-reported "fatigue/fogginess." Cognitive deficits using the Digit Symbol Substitution Test and Trail-Making Test-part B recovered concomitantly with symptoms, but computerized test results recovered 2 to 3 days later and remained impaired in 35% of concussed players after symptom resolution. CONCLUSION: Delayed return to sport was associated with initially greater symptom load, prolonged headache, or subjective concentration deficits. Cognitive testing recovery varied, taking 2 to 3 days longer for computerized tests, suggesting greater sensitivity to impairment. Therefore, symptom assessment alone may be predictive of but may underestimate time to complete recovery, which may be better estimated with computerized cognitive testing.