News
COLUMN – Science BEAT

by Brent Snook

Welcome to the inaugural Science Heat, a column which investigates the hottest scientific findings that matter to law enforcement officers. The goal is to provide readers with a review of current research on policing issues and advance scientific literacy. Although developments from psychological research will be the focus, key findings from other scientific disciplines will also be highlighted.

Deception detection has dominated the pages of forensic psychology journals recently and much is being written about how well deception detection tools work. Nahari Galit (naharig@mail.biu.ac.il), Aldert Vrij and Ronald Fisher estimated recently that Scientific Content Analysis (SCAN) is the most common method used by police worldwide to detect lies.

The SCAN technique was developed by former Israeli polygraphist Avinoam Sapir as a way to measure the truthfulness of verbatim statements. Examinees are required to write a detailed description of all their activities during the period when the crime occurred. The handwritten statement is then examined for the presence of criteria meant to help decide if it is truthful (e.g., direct denial of the allegation in the statement) or deceptive (e.g., deviation from telling a story in chronological order). Although SCAN has been around for a long time, only a handful of studies have tested how well it works.

In a recent study published in (a journal that publishes high-quality research on forensic psychology issues) Galit and colleagues tested how well SCAN could detect verbal lies of 61 undergraduate students. It was also compared to an alternative method known as Reality Monitoring (RM), which is based on the idea that real memories of experienced events are externally derived and based on actual experiences (e.g., where the event took place, what the person heard, emotions). Fake memories, however, are internally derived and based on self-generated thought or imagination (e.g., no spatial details provided in the account). In essence, the RM technique assumes that a statement is truthful if there is evidence of real experiences and false if there is evidence of imagined experiences (or a lack of real experiences).

In their study, students were assigned randomly to do one of three things: (a) lie about a criminal activity but tell the truth about non-criminal activities, (b) lie about both criminal and non-criminal activities and (c) tell the truth about all activities. The two “lie” groups (a and b above) were asked to follow an elaborate set of instructions about how they should commit a mock theft and engage in non-criminal activities (e.g., converse with a friend for 30 minutes). Those in the truthful condition were asked to carry out a variety of non-criminal activities for the same period of time. All participants were then motivated to be as convincing as possible that they were telling the truth, asked to write down as much detail as possible about what they were doing during the 30 minutes and asked not to change anything they wrote.

Six trained coders then checked the student’s statements for the presence of 13 SCAN criteria and eight RM criteria without talking to each other. The coders were also not told about the purpose of the study to prevent any potential biases from creeping into the findings. Approximately half of the criteria had 80 per cent agreement or greater between coders.

Their results showed that the single SCAN score (from summing the 13 SCAN criteria) was unable to identify the students who were lying. None of the criteria (even those that had very high levels of agreement between coders) was helpful in identifying truthful or deceptive statements. Deception detection accuracy hovered around chance levels (50 per cent) or just above chance (60 per cent).

Interestingly, Galit and colleagues found that there were more RM criteria present in the truthful statements than the deceitful ones. The RM technique was approximately 71 per cent accurate in classifying statements as truthful or deceptive (i.e., those that lied about all activities).

Because the coders disagreed about whether some criteria was present in a statement, the amount of trust we can put into the results is limited. Since about half of the SCAN criteria are highly subjective, two different people could provide two different answers about deception when looking at the same statement.

Given that there have only been a handful of studies on this topic and the reliability of the data in the study reviewed wasn’t overly high, it’s especially important that other scientists repeat the study to see if they can get higher agreement levels for the data and determine if the reported results can be found again (i.e., replication). It might be the case (if we had highly reliable data) that RM is a viable alternative for detecting lies. What we do know, however, is that there is currently no compelling empirical data to support SCAN as a useful way of detecting deception.

BIO BOX

Brent Snook, B.A., M.Sc., Ph.D., is an associate professor in the Psychology branch of Memorial University in St. Johns Newfoundland. Contact him at bsnook@mun.ca or 709 864-3101 for more information.