Deceptive dangers of lie detectors

Britain is likely to mandate from next January the use of the polygraph to monitor sex offenders who are released after serving their sentences.

But its questionable reliability means that polygraph evidence remains inadmissible in British courts, as it also is in Singapore. Here, law enforcement officers have nevertheless resorted to the instrument since 1977 to assess the credibility of suspects and help decide what charges to file, if any.

In some civil cases, people submit reports of polygraph tests they have taken on their own accord, there being a company offering the test here that was set up by a former police officer in 2001.

But the polygraph can't be described as an extremely accurate lie detector. This is because the physiological responses measured don't, in sum, point to a single underlying process of deception.

The instrument measures the examinee's breathing rate, pulse rate, blood pressure and galvanic skin reflex. The last item is based on the fact that moist palms conduct more electricity than dry ones.

The idea is that, on being asked a question to which he plans to lie, the examinee's brain sets off automatic physiological changes that he can't control which the polygraph inexorably captures.

But lying doesn't evoke a unique set of physiological responses. A guilty person's fear of being caught may lead to faster breathing (monitored with belts around the chest and tummy), a quicker pulse and elevated blood pressure (monitored with a blood pressure meter cuff around an arm) and raised skin conductance (monitored at the fingertips).

But a non-guilty examinee's fear of a "false positive" can evoke the same responses too. Nervousness, embarrassment or outrage may cause the same responses as well. Since these alternatives can also explain these measurements in each case apart from lying, what proponents need to establish is a logical explanation or "theory" that explains the link between these physiological responses and guilt (deception).

Such a theory would consist of specific factors or mechanisms that work together coherently to explain that link. Then research must fine-tune the science in that theory as new evidence accumulates. This is how all conventional scientific fields advance.

The problem with polygraph research is that it has never formulated such a theory. In all these decades, polygraph research has been atheoretical, meaning that it has never been directed towards proving how specific factors and mechanisms link particular physiological responses to actual lying.

Instead, polygraph research only tries to prove it "works" under lab conditions, where a person is giving facts from which he is to lie when tested. Other possible causes of any spikes are not excluded.

There is no reason to assume that the polygraph works exactly like that in real life since life does not replicate artificial lab conditions. Also, no randomised, controlled trials (RCTs) with appropriate placebo groups have ever been done.

The United States National Research Council (NRC) conducted a comprehensive review of polygraph research in 2003 which found none had developed a theory why the polygraph should work. It concluded that the polygraph's reliability was questionable, and neither technological nor methodological advancement was likely to improve that reliability.

A decade on, the NRC review is still definitive as polygraph research has remained atheoretical. No RCTs have been conducted, so the field has not advanced scientifically at all.

Most of these newer studies are confounded by biases because the most critical part of the test is not the machine but the questions polygraphers ask. Yet their training is not rigorous. The oldest polygrapher trainer, America's 60-year-old National Centre for Credibility Assessment, takes only 14 weeks to train a polygrapher.

There are no standard protocols of questions to ask. Instead, the polygrapher formulates them as he goes along. By contrast, two well-trained X-ray technicians using the same machine and protocol to image a patient's chest will produce the same chest X-ray picture. With the polygraph, however, the results are highly dependent on the procedure, there being no standard set of questions to ask since these would depend on what the case involves.

An offender's unknown sex crimes exacerbate the difficulty of formulating appropriate questions. Moreover, examinees can learn to control the physiological responses that lying supposedly causes. The NRC report called such polygraph counter-measures "learnable". A Google search for "beating" or "cheating" the polygraph yields 180,000 returns. These include doing mental arithmetic or biting the tongue hard or contracting the anal sphincter.

On pre-test questions used to establish a baseline - "Have you ever cheated on a test?" or "Have you ever lied to your spouse?", say - you answer "No" and bite your tongue hard, the pain causing your physiological responses to spike. Knowing that most of us would have done these things at least once, the tester thinks you might well be a nervous liar and easy meat.

But then any spikes when relevant questions about an actual offence are asked would not be as big by comparison because you are not biting your tongue hard, so the spikes caused when you really lie are less. You won't be seen as lying, so you beat the test.

Since its reliability is questionable and, being a test that can be beaten, it seems best not to allow it to go too far into the criminal justice system. The NRC review also found it inappropriate "for test results to play any role in determining whether an individual will be deprived of personal liberty". Using the polygraph to assess examinee credibility in a criminal justice context is unwise.

Instead, its use might best be limited to a purely therapeutic context. By de-linking it from the threat of punishment, the sex offender will have less reason to adopt counter-measures, which might make the test more reliable and helpful in his rehabilitation.

Get a copy of The Straits Times or go to for more stories.