A review of the evidence now accumulated about these devices shows that the evidence for the existence of a microtremor in the voice is problematic and that the capability of these devices in detecting stress is equally questionable. Without exception, however, the scientific evidence reported to date shows that voice stress analyzers are not effective in detecting deception; none of these devices has yet been shown to yield detection rates above chance levels in controlled situations. A brief comparison of voice stress analysis and polygraphic testing as methods of lie detection is made. (Author abstract)