Concussion, or mild traumatic brain injury, occurs in many activities, mostly as a result of the head being accelerated. A comprehensive study has been conducted to understand better the mechanics of the impacts associated with concussion in American football. This study involves a sequence of techniques to analyse and reconstruct many different head impact scenarios. It is important to understand the validity and accuracy of these techniques in order to be able to use the results of the study to improve helmets and helmet standards. Two major categories of potential errors have been investigated. The first category concerns error sources specific to the use of crash test dummy instrumentation (accelerometers) and associated data processing techniques. These are relied upon to establish both linear and angular head acceleration responses. The second category concerns the use of broadcast video data and crash test dummy head-neck-torso systems. These are used to replicate the complex head impact scenarios of whole body collisions that occur on the football field between two living human beings. All acceleration measurement and processing techniques were based on well-established practices and standards. These proved to be reliable and reproducible. Potential errors in the linear accelerations due to electrical or mechanical noise did not exceed 2% for the three different noise sources investigated. Potential errors in the angular accelerations due to noise could be as high as 6.7%, due to error accumulation of multiple linear acceleration measurements. The potential error in the relative impact velocity between colliding heads could be as high as 11%, and was found to be the largest error source in the sequence of techniques to reconstruct the game impacts. Full-scale experiments with complete crash test dummies in staged head impacts showed maximum errors of 17% for resultant linear accelerations and 25% for resultant angular accelerations.