Voice Stress Analysis

Voice stress analysis (VSA) was developed in the 1970’s by three American army personnel who were first tasked with enabling voice identification of intercepted radio communications. The premise is that changes in something known as the physiological microtremor due to the fight or flight response will be noticeable in graphic representations of a subject’s answers. It was first accepted as a mean of psychophysiological detection of deception however many studies since have concluded that it is unacceptable, yielding results ‘no better than flipping a coin.’

There are several issues with VSA apart from these results. The physiological microtremor is present in all mammals at normal frequencies of 8-12Hz and under physiological stress is reduced to 4-6Hz. No microphone available to the consumer is capable of picking up these frequencies with most microphones available only having frequency ranges of 100Hz – 10kHz. Even the better ones do not go below 20Hz. Therefore these microphones are not collecting the correct frequencies to even display the physiological microtremor within an answer’s representation. Also, the only physiological reactions being recorded are at the moment of the subject’s answer which is a maximum of one second. Although a physiological response is generally initiated within milliseconds of a perceived threat they are different for each individual and can last for up to 15 seconds. Thus VSA is taking only a tiny portion of the response (if any) to be analysed, as opposed to traditional polygraph which is continuously measuring. This can lead to serious errors.

The three largest associations; the American polygraph association (APA), the European polygraph association (EPA) and the Israeli computerised polygraph association (ICPA) all do NOT condone Voice Stress Analysis (VSA) in any form.

There are also programs available for VSA claiming autonomous scoring. These have been thoroughly tested by our institute and proven faulty. A single recording of one answer was played repeatedly directly into the system and the values returned differed each and every time. Definitely not what one would expect.

All in all, VSA still has potential despite it’s flaws but a lot more research and development needs to be done before it can be truly accepted as a justifiable detection of deception technique.