In the game Dungeons & Dragons, there is a magic spell called detect lie (or at least there was back in the first edition). A couple of my favorite high school gaming sessions revolved around the player characters flinging accusations at each other - and immediately finding out which one was lying.
According to psychologist Aldert Vrij's excellent book Detecting Lies and Deceit, those of us without magical powers have a long way to go. Not only are laymen bad at detecting lies; so are policemen, customs agents, and other people who do it professionally.
How bad? There are many experiments where subjects are required to lie and tell the truth with equal probability. Observers then try to sort fact from fiction. A representative result: People correctly identify truths 70% of the time, but correctly identify lies only 50% of the time. If you know Bayes' Rule, you can use this information to calculate the probability a statement is true given that it seems true:
which by my calculations=58.3%. A little better than random guessing, but not much.
A few other juicy morsels from Vrij:
"[O]bservers overestimate the likelihood of being able to detect deceit by paying attention to someone's behavior, and... underestimate the possibility of catching liars by paying attention to their speech content."
Contrary to popular opinion, gaze aversion does not predict lying. This may be because almost everyone believes it does, leading even inexperienced liars to try not to avert their gaze.
It is easier to tell if someone is lying if you are familiar with their ordinary (non-lying) speech and behavior. But actually meeting or intimately knowing the suspected liar does not give an additional benefit. In other words, you are more able to detect lies in your friends because you know how they normally act, not because people look guiltier when they lie to their friends.
Men and women lie equally often, but men tell more self-oriented lies and women tell more "other-oriented lies, particularly with regard to other women." (Self-oriented lies are designed to gain an advantage for the liar; other-oriented lies are designed to help someone else - usually the listener).
Call me a liar, but Vrij's book convinced me that I am almost pathologically honest. I would certainly lie to save an innocent person's life. But the common sense moral truism that it is wrong to lie still seems compelling to me, and I adhere to it. (Yes, I just averted my gaze from the monitor, but that proves nothing!) Vrij argues that lies lubricate social relations, but there are honest ways to do the same thing. Most social pleasantries are non-propositional anyway; if someone says "Thank you," you cannot coherently respond "False!"