Review of Voice Stress Based Technologies For the Detection of Deception

Executive Summary

A review of 24 empirical studied that were conducted over a period of more than 30 years failed to provide evidence of the validity or reliability of voice stress analysis-based technologies for the detection of deception in individuals. There was some evidence that some voice analysis-based technologies are able to detect voice stress reliably. A more detailed description of two recent empirical studies, both exemplary for providing independent assessments of stress and deception (Hollien, Harnsberger, Martin, and Hollien, 2008) and deception (Damphouse Pointon, Upchurch, and Moore, 2007) are described in greater detail so that some of the requisites for a good empirical study can be understood.

We had access only to website descriptions provided by voice stress analysis-based technology vendors and one field survey (conducted by DoD) for evaluation of the narratives (anecdotes) provided by field use of these technologies. While a great many of these anecdotes are likely to reflect various, commonly occurring cognitive heuristics that result in non-rational judgments, we recomend that these field reports not be ignored but that they be subjected to a more rigorous analysis. The full report can be downloaded here: