Concussions occur at an unacceptable rate in the sport of ice hockey. Efforts are made to improve its prevention by modifying protective equipment and implementing rules of conduct; yet the effectiveness of these methods remains unknown as there is a lack of evidence pointing to a mechanical metric able to adequately predict concussion. The purpose of this thesis was to identify metrics which best characterize concussion following ice hockey collisions and provide values reflecting concussion risk. The first study reported effective mass of shoulder checks, extended elbow strikes, and tucked-in elbow strikes using 15 competitive ice hockey players as subjects. The results were used to guide the impact mass and compliance of laboratory reconstructions of real-world ice hockey collisions done in the second study. Analysis of these reconstruction showed that concussions following shoulder and elbow to head collisions in ice hockey occurred at low peak linear and angular accelerations and that impulse duration played a large role in the mechanism of injury. The results also indicated that concussion risk estimations are specific to the mechanism of loading. A 50% likelihood of concussion following a shoulder check to the head was established for peak angular accelerations of 9.2, 6.9, 4.6, and 2.2 krad/s^2 for impulse durations of 15, 20, 25, and 30 ms, respectively. A 50% likelihood of concussion following an extended elbow to the head was established for peak linear accelerations of 23, 15, and 7 g for impulse durations of 15, 20, and 25 ms, respectively. Finally, the third study reported brain tissue stress and strain comparable to the ones obtained reconstructing concussive impacts in American football, rugby, and Australian rules football despite having lower peak linear and angular acceleration values. This thesis has provided a new sport concussion data set acquired using a methodology guided by the biomechanics of ice hockey player volunteer testing, has identified metrics which can adequately predict concussion, and has established concussion risk levels. This information will be of use to helmet manufacturing companies, companies developing concussion detection sensors, and governing bodies in their efforts to eliminate concussion from the sport of ice hockey.