>So how can you know how to cheat them? I don't understand the basis of your question. To say "this test does not correlate with the phenomenon it purports to

>So how can you know how to cheat them? I don't understand the basis of your question. To say "this test does not correlate with the phenomenon it purports to measure" is not the same as to say "this test is completely random". Counting how many times a person says the word "baby" during an answer is a piss-poor way to tell if they're lying, in large part because that factor won't correlate strongly with deception. It would also be extremely easy to fool, by simply controlling how many times you say "baby".