You can now find the University's latest news, events and blogs at

‘Smart’ liars caught out by researchers

Lie detector Dom Shaw

Lie detector Dom Shaw

Researchers at the University of Portsmouth may have found a way to catch out liars who pass an established lie detection test.

It is the first experiment to reveal a pattern in how guilty people approach what’s known as a Symptom Validity Test.

The test is a clinical assessment tool used by neuro-psychologists to determine if someone’s claims of amnesia are genuine, and it bears no resemblance to better known lie detection tests, such as the polygraph. The researchers adapted the test to see whether participants would incriminate themselves by avoiding guilty knowledge.

In the experiment 86 people were asked to commit a mock crime by stealing a data in the form of a PDF named ‘delta’, from a computer with a distinct yellow background.

A further 82 people allocated to the control group had no knowledge of the crime. All were then submitted to a lie detection test that looks for signs that someone is avoiding associating themselves with anything directly related to the ‘crime’.

The Symptom Validity Test requires people to choose one word from a pair of words. Each pair contains a word relating to the crime and a neutral word not related to the crime. In this experiment, crime-related words included ‘PDF’, ‘delta’ and ‘yellow’ and neutral words included ‘JPEG’, ‘bravo’ and ‘red’.

In the test, innocent people with no knowledge of the crime chose an equal number of crime-related and neutral words.

Liars, on the other hand, chose the crime-related words just 33 per cent of the time and 40 per cent scored below chance and failed the test after choosing three or fewer crime-related words.

If someone fails the test, it is a strong indication of guilt. However, the researchers caution that passing the test does not indicate innocence, and it is possible to beat the test by using a strategy.

Researcher Dom Shaw said: “What is unique about this study is that we found a pattern in the guilty participants’ responses.

“Our results suggest that at some point during the test, some liars worked out that avoiding too many crime-related words would appear suspicious. As a result, they started including more crime-related words to appear as if they were choosing purely by chance.”

Some guilty participants who beat the test tended to avoid the crime-related words at the beginning, but then included some later on, indicating they understood how the test works and that avoiding the guilty knowledge too often will appear deceptive.

These ‘smart’ liars may have passed the test when looking at their overall scores, but they still left evidence of their guilt in their pattern of responses, said Mr Shaw, and this information might increase the value of the test because passing it in this way might indicate guilt.

After the experiment, participants were asked what strategies they had used to answer the test. The two most popular strategies cited by the liars were to deliberately select some correct answers and attempt to randomise their answers.  Of those who expressed some understanding of how the test works, a third still failed.

Mr Shaw said: “Some liars may have grasped how the test works too late, meaning they simply didn’t have enough time to perform at chance level.

“One other explanation is that deceptive participants simply struggled to answer randomly. While it is not clear exactly why this might be the case, there is some suggestion that humans struggle to make random responses.”

The researchers suggested that a test containing more than 12 questions might reveal more about the liars’ deceptive response patterns.

Mr Shaw said: “With a longer Symptom Validity Test we might see liars switching back and forth between selecting each type of word for a period of time. In their attempts to answer randomly, liars may in fact incriminate themselves.”

The researchers said that while the findings are encouraging, the results need to be treated with caution.

Mr Shaw said: “The test shows promise in detecting feigned crime amnesia. Our findings should be treated as a foundation, and future research is needed.

To further explore the value of the findings, the next step for the researchers is to examine data from those who have been coached to beat the test.

Mr Shaw said: “Our participants were not told how they could beat the test, but in real life, barristers may coach their clients about how to pass neurological tests. The next challenge is to examine the response patterns of those who have been coached to see if this reveals similar evidence of deception.”

The research is published in Legal and Criminological Psychology.

3 total comments on this postSubmit yours
  1. “Lying is a skill like any other and if you want to maintain a level of excellence, you have to practice constantly.” – Elim Garak

    One experiment measured the brain structure of pathological liars, and compared it to normal controls — more specifically, the ratio of gray matter (the neural tissue that makes up the bulk of our brains) to white matter (the wiring that connects those brain cells). Liars, it turned out, had 14% less gray matter than the controls but had 22-26% more white matter in the prefrontal cortex, suggesting that they were more likely to make connections between different memories and ideas as increased connectivity means greater access to the reserve of associations and memories stored in gray matter.

  2. This is not correct. This article is a lie. And you don’t need a credibility assessment tool to understand that. The test invented at University of Portsmouth is not near as accurate as a properly administered polygraph test. Nor has it been investigated whether it actually can improve accuracy in combination with a polygraph test. The superior utility of this new test, claimed by the author of the above article is false.

    National Research Counsil in the US found in 2003 that the median accuracy for polygraph tests when using the 30 best research papers (not with the highest accuracy, but papers with best research quality) was 85%. When using the 7 studies with the highest research quality, the mean accuracy was 89%. Now… most people stop to think there and believe that is the answer and that this is how accurate polygraph tests are. I’ve even seen several psychology professors go in that trap. Everyone who read this, should know from now on, that 85% or 89% is not the accuracy of polygraph tests.

    There are a lot of different polygraph methods. Probably somewhere between 70-100, if you include variants of different techniques. Only some of these techniques has beed subject to research, and unfortunately, much of the research which has been done are of poor research quality. Some of these methods are similar in design and scoring methods, others are significantly different. If you start changing the way you ask the questions, which type of question you use in a test, and score tests with totally different methods, I hope everyone who reads this post, understand such changes will also affect accuracy.

    Just to give you an idea: one method have no structured pre-test interview, and the strongest relevant question is put before any comparison questions. This will increase the bias against the innocent, and the effect is strong. This method has also got some rules for how to construct comparison questions which makes it less standardised and difficult to know if the comparison question used is a strong or weak question. Further, this method has a scoring system which is completely inferior to the best scoring systems available today. And last: this method has just three types of questions, irrelevant, comparison and relevant questions, which means the test does not cover possible errors that are covered by the best methods available today, which use several other questions to deal with for example outside issue related problems and countermeasures. One of these simple tests has shown an average accuracy in the range of 60-70% with, in one study, as low as 25% detection rate for the innocent (other studies have shown around 40-50% accuracy for the innocent). Unfortunately, some polygraph examiners, are at least as ignorant as the person who wrote the above article, and have no idea that this type of polygraph technique is totally useless and extremely dangerous to use. Sadly this method is still used by some polygraph examiners, and the examiners have no idea what they are doing.

    On the other hand, we have today 9 polygraph methods with a proven accuracy above 90% for single issue testing. 5 of them range from 90-95% accuracy, and 2 has an average accuracy of 99.4%. The two last methods are very similar in test design, but are scored differently, and they account for 24.8% of all tests in studies which are been labeled to be of good quality.

    The tests done in studies on the two best polygraph formats, are mostly field studies with real cases. These are real guilty people who tries their best to pass the test. And they are real innocent people which are tested for real crimes which involves high stakes if caught, and quite possibly also a lot of emotions. Even so, the polygraph tests come out with a very high accuracy rate.

    It’s great that new non-polygraph methods is invented for the detection of deception. We need more of those, but there is no need to falsely inflate their accuracy or utility.

  3. May I add that National Research Counsil mixed several types of methods when they claimed a median 85 to 89% accuracy for single issue polygraph tests. That way of calculating accuracy shows total lack of knowledge about polygraph tests (the researchers were not polygraph examiners) and would be like mixing apples, tomatoes, coffee and kaviar, and afterwards state that apples aren’t as good as apple eaters claim, and is actually junk food.
    Eveyone, even thos who don’t eat apples can understand how silly that is.

    Of the 7 best research studies, the accuracy of the different polygraph techniques and studies, varied from 71.1% to 99.9%. The study with 99.9% was the first study of one of those techniques which now hold a 99.4% accuracy. Since 2003 several studies has been done which confirm the high accuracy of this technique, and another technique with only slight difference in design also has 99.4% accuracy.

1 pingback on this post

UoP News © 2019 All Rights Reserved