Review of a large data base on 76 patients who had suffered head injuries and whose metabolic changes were being studied demonstrated the variability of their neurological courses and the importance of considering confounding factors in the analysis of such a group. It was apparent that many such patients continue to change neurologically and functionally for long periods of time, and that their ultimate outcome cannot be determined for at least 1 year. The study demonstrates the inherent complexity of a group of patients with head injuries and the need to use sophisticated statistical tools and a computer to demonstrate whether apparent trends are statistically valid in light of confounding factors and variabilities. Failure to take this complexity into account seems to be a major factor in many of the ongoing controversies in research on head injuries. A possible solution to the problem of potential biases introduced by lost patients would be to develop a method to use government records to obtain current addresses for patients involved in federally funded research.