You need to take the guesswork out of the equation, and filter out all of those global behaviors that do nothing to help you get to where you want to go: identifying deception. So, you need to limit your analysis to only those behaviors that come in direct, timely response to the stimulus, which is your question.
