This paper is about the major peculiarities and difficulties we encounter when trying to validate research results in fields such as virtual reality (VR) and 3D user interfaces (3DUI). We review the steps in the empirical method and discuss a number of challenges when conducting human-subject experiments. These challenges include the number of independent variables to control to get useful findings, the within-subjects or between-subjects dilemma, hard-to-collect data, experimenter effects, ethical issues, and the lack of background in the community for proper statistical analysis and interpretation of the results. We show that experiments involving human-subjects hinder the adoption of traditional experimental principles (comparison, repeatability, reproducibility, justification and explanation) and propose some ideas to improve the reliability of findings in VR and 3DUI disciplines.