Seeing through the face of deception

There is an urgent need to devise technologies that can be used for automated, high-throughput screening to identify individuals intending to perform acts of terrorism. At present, practicalities dictate that we rely on subjective assessment of responses to brief questions such as “Did you pack your own bags?” and “Why are you entering this facility?”

Although polygraph examinations, which have high precision when applied by experts¹, are good at identifying liars, they are impracticable for mass screening because skilled operators are needed, subjects have to be attached to instrumentation for several minutes, data analysis is time-consuming and the interpretation of data is delayed.

We explored the possibility of using high-definition thermal imaging of the face for detecting deceit² because it enables rapid automated analysis of changes in regional facial blood flow to be quantified^3,4. We have shown previously² that auditory startling is associated with a specific facial 'thermal signature', in which there is instantaneous warming around the eyes — probably as part of a fright/flight response mediated by the sympathetic nervous system^5,6. Although the psychophysiology of startling differs from volitional deception, the nonspecificity of this facial thermal signature is reminiscent of the nonspecific variables monitored during a polygraph (respiration, pulse, relative blood pressure and electrodermal response). Were this thermal signature to accompany lying, independently of startling, it could be used for instantaneous lie detection without the subject even being aware of the test.

We therefore asked volunteers to commit a mock crime and then testify to their innocence under experimental conditions at the US Department of Defense Polygraph Institute (DoDPI; http://www.dodpi.army.mil)⁷. Twenty individuals were randomly assigned to stab a mannequin, rob it of $20 and then assert their innocence of the 'crime'. Control subjects had no knowledge of the crime or of the crime scene. The thermal imaging system correctly categorized 83% of these subjects (Fig. 1); three-quarters (6 of 8) of the guilty individuals were correctly identified as guilty and 90% (11 of 12) of the innocent individuals were correctly categorized as innocent. Traditional polygraphs, performed by experts at DoDPI on the same subjects, correctly categorized 70% of the subjects: 6 of 8 subjects were correctly identified as guilty and 8 of 12 were correctly identified as innocent. Under these experimental conditions, the accuracy of the thermal imaging system was comparable to that of the traditional polygraph.

Images were obtained before (a) and after (b) lying in reply to the question “Did you steal the $20?” Images were obtained at 30 frames per second with a cooled thermal camera with a thermal sensitivity of 0.025 °C. The camera was calibrated daily to T_min = 29.00 °C (black) and T_max = 38.00 °C (cyan) with an external black body; red, orange and yellow represent progressively warmer temperatures in between. White lines indicate eye contours.

Full size image

High-definition thermal imaging of the face is therefore a promising technology that should allow psychological responses to be detected and analysed rapidly and without physical contact, in the absence of trained staff and in a variety of different situations.