There are many basic experiments that could be designed to at least narrow down the reliability of the data. I could imagine a fairly simple test in which actors are hired to answer scripted questions with scripted answers, they are video recorded wearing suit A and then suit B. You'd use a large set of actors to help…