Data on helmet models used and occurrence of cerebral concussions over five seasons were collected from a representative sample of college football teams including a total of 8,312 player-seasons and 618,596 athlete-exposures to the possibility of being injured in a game or practice. Results showed that players with a history of concussion any time during the previous 5 years were six times as likely to suffer a new concussion as those with no previous history. In light of previous studies showing cognitive deficits for up to 30 days following even minor head injuries, and the growing awareness of "second impact" fatalities, these data support a need for reconsideration of the common practice of immediate return to play following non-loss-of-consciousness head injuries. Results on concussion frequency in ten models of football helmets indicated a significantly lower than expected frequency in the Riddell M155 and a significantly higher frequency in the Bike Air Power. All other models performed within expectations. This study demonstrates the need for monitoring on-the-field performance of football helmets through continuing epidemiological studies to supplement laboratory test data, which cannot duplicate all the factors involved in actual helmet performance.