Meet and Beat the Lie Detector
by William Poundstone
The polygraph test was invented by William Moulton Marston, who was,
strangely enough, also the creator of the "Wonder Woman" comic strip
(under the name Charles Moulton). The standard polygraph records
only three distinct vital signs. A blood-pressure cuff on the upper
arm measures changes in blood pressure. Wires attached to the
fingers measure changes in electrical resistance of the skin due to
sweating. Rubber straps around the torso measure the breathing
rate. This information is displayed as four squiggles on a moving
strip of graph paper.
Whether or not you believe a polygraph provides useful information
(most psychologists have their doubts), there is a good chance you'll
be asked to take a polygraph test. The vast majority of lie-detector
tests are administered for employee screening--"Have you been using
the WATS like for personal calls?" and so forth--not for police
work. In 'A Tremor In the Blood: Uses and Abuses of the Lie
Detector' (New York: McGraw-Hill, 1981), polygraph critic David
Thoreson Lykken estimates that as many as one million polygraph
examinations are performed on Americans each year. In criminal cases
however, even the manifestly innocent may be asked to take a
polygraph test. All Yakima County, Washington, rape victims are
required to take the test; refusal means the case will not be
prosecuted.
At best, all the polygraph can indicate is a heightened emotional
reaction to a question. It cannot specify what kind of an emotional
reaction. Polygraphers try to design question formats so
guilt-induced nervousness will be the only emotional invoked and so
the subject's reaction to relevant questions can be compared to
other, "control" questions.
THE LIE CONTROL TEST
This is the question format used in most police investigations. It
usually starts with a card trick devised by two pioneer polygraphers,
John E. Reid and F. E. Inbau.
The polygrapher hooks the subject to the polygraph and takes out a
deck of cards. The polygrapher tells the subject that he must
"calibrate" the polygraph with a simple test. He fans the deck and
asks the subject to select a card. The subject is told to look at
the card but not to show it or mention its name. The polygrapher
tells the subject to answer "no" to every question asked about the
card. "Is it a black card?" the polygrapher asks. "Is it a high
card?" and so on. After each "no" the polygrapher scrutinizes the
tracings and fiddles with the dials. If the no answer is incorrect,
the polygrapher disagrees. The field is soon narrowed to one
card--and it is the correct card.
Needless to say, the polygrapher uses a trick deck. The point is to
foster confidence in the machine. After identifying the card, the
polygrapher comments that the subject's reactions are particularly
easy to read and segues into the interrogation.
Three types of questions are used in a lie-control test. The entire
list is read to the subject well in advance of the test. The start
of a typical interrogation might run like this:
1. Is your name Sarah Elkins?
2. Is Paris the capital of France?
3. Have you ever failed to report more than $50 of tip, gambling or gift
income on a single year's tax return?
4. Is this apple red?
5. Do you have any idea why the cash reciepts for the last quarter are about
$22,000 in error?
6. Is there something important that you did not mention on your job
application?
7. Have you ever been embezzling from the company?
The first question is always irrelevant to the matter being
investigated. It has to be because many subjects get nervous on the
first question no matter what. Other irrelevant questions are asked
throughout the interrogation (questions 2 and 4 in the sample list).
If the subject gives any thought to these questions, he assumes that
they are control questions to provide a yardstick for evaluating
responses to the relevant questions. Actually, the irrelevant
questions are there to give the subject's vital signs time to return
to normal. They aren't the control questions.
Questions 5 and 7 in the list above are relevant questions--the only
questions the examiner is really interested in. The relevant
questions are asked in several different wordings during the test.
Questions 3 and 6 are control questions. In the pretest discussion
of the questions, the polygrapher explains that it is helpful to
throw in a few "general honesty" questions. Whoever committed the
serious crime, the spiel goes, probably comitted less serious crimes
in the past. Hence the inclusion of questions about tax cheating,
lying on the job applications, stealing as a child, etc.
The polygrapher affects the attitude that it would be damaging indeed
to admit any such indiscretions. Frequently this scares the subject
into admitting minor crimes. In that case, the plygrapher frowns and
agrees to rewrite the question. Should the subject concede failing
to report eighty dollars in gambling winnings, question 3 might be
changes to "Have you ever failed to report more than a hundred
dollars of tip, gambling, or gift income on a single years's tax
return?" If necessary, several of the control questions may be
reworded before the test--always so that the subject will be able to
give the "honest" response.
In reality, the whole point of each working question is to
manufacture a lie. It is the secret working premise of polygraphers
that everyone commits the minor transgressions that are the subject
of the usual control questions. All the subject's denials on the
control questions are assumed to be lies. The polygraph tracings
during these "lies" establish a base line for interpreting the
reaction to the relevant questions.
The reason for rewriting some control questions is so a candid
subject will no admit to minor crimes on the test. That would be
telling the truth, and the polygrapher wants the subject to lie. The
control questions are intentionally broad. Even if a question is
reworded to exclude the confessed instance, it is assumed that any
denial must be a lie.
The rationale for the lie-control test goes like this: The honest
subject will be worried about the control questions. He'll know that
he has committted small transgressions or suspect that he must have,
even if he can't remember them. So he'll be afraid that the machine
will detect his deception on the "general honesty" questions
(especially in view of its success with the card trick). That would
be embarrassing at least, and it might throw suspicion on him for the
larger crime. In contrast, the relevant questions should be less
threating to the honest subject. He knows he didn't commit the
crimes they refer to.
The guilty person, on the other hand, should have far more to fear
from the relevant questions. If the machine can detect lying on the
relevant issue, it matters little that it might also implicate him in
petty matters.
By this hypothesis, an innocent person should have greater
polygraphic response to the control questions than to the relevant
questions. The guilty pattern is just the reverse: greater response
to the relevant qustions. This, at any rate, is what polygraphers
look for when the machine is switched on.
THE RELEVANT-CONTROL TEST
The relevant-control test is the type used for most employee
screenings. Thus it is the most common type of examination. The
interrogation consists only of irrelevant and relevant questions. As
with the lie-control test, the first question and a few others are
irrelevant. The relevant questions usually test workplace honesty:
"Have you ever taken home office supplies for personal use?" "Have
you ever clocked in for someone else?"
The premise is that no one will lie about everything. So if a few of
the relevant questions produce heightened responses, they are
presumed to be the questions on which the subject is lying.
Unfurtunately, there is no ambiguous way of deciding how much
response is indicates a lie. Most psychologists agree that the
relevant-control test is a poor test of deception.
The Reid/Inbau card trick is eliminated from employee screenings:
There is too great of a chance of coworkers comparing notes and
discoverings that everyone picked the ace of spades.
HOW TO BEAT THE LIE-DETECTOR
To the extent that the polygraph works at all, it works because
people believe it does. Many criminals confess during polygraph
examinations. Many employees are more honest for fear of periodic
screenings. But a dummy polygraph that hummed and scribbled
preprogrammed tracings would be no less effective in these instances.
David Thoreson Lykken estimates that lie-control polygraph tests are
about 70 percent accurate. (Remember, though, that choosing "heads"
or "tails" of a flipped coin can be accurate 50 percent of the
time.) Accuracy of 70 percent is not impressive, but it is high
enough to talk meaningfully of beating a polygraph test.
Just by having read this far, you stand a greater chance of beating a
polygraph test. You won't be wowed by the card demonstration. You
realize that the polygraph's powers are limited. There are two
additional techniques for beating the polygraph. The more obvious is
to learn how to repress physiologic responses to stressful
questions. Some people are good at this one; others are not. Most
people can get better by practicing with a polygraph. Of course,
this training requires a polygraph, and polygraphs are expensive.
The opposite approach is to pick out the control questions in the
pretest discussion and exaggerate reactions to these questions during
the test. When the control-question responses are greater than the
relevant-question responses, the polygrapher must acquit the subject.
Because breathing is one of the parameters measured, taking a deep
breath and holding it will record as an abnormal response. Flexing
the arm muscles under the cuff distorts the blood-pressure reading.
But a suspicious polygrapher may spot either ruse.
A more subtle method is to hide a tack in one shoe. Stepping on the
tack during the control questions produces stress reactions with no
outward signs of fidgeting. Biting the tounge forcefully also works.
|