The detection of deception by linguistic means Unresolved issues of validity, usefulness and epistemology

摘要

Purpose - The purpose of this paper is to discuss unresolved problems that are reflected in the social scientific research on the linguistic detection of deception in statements, with particular attention to problems of methodology, practical utility for law enforcement statement analysts, and epistemology. Design/methodology/approach - The author reviewed the design, data, statistical calculations, and findings of English language peer-reviewed studies of the linguistic detection of deception in statements. In some cases, the author re-analyzed the study data. Findings - Social scientific research holds promise for the development of new methods of linguistic detection of deception that are more thoroughly validated than the linguistic methods law enforcement investigators have been using for many years. Nonetheless, published studies reflect one or more of the following sources of weakness in developing and evaluating detection models: the use of analytes (statements) of uncertain validity; the problematic universality and practical utility of linguistic variables; the widespread use of deficient proportion-of-stimuli-correct "hit rate" calculations to assess the accuracy of detection methods; a possibly irresolvable epistemological limit to the ability of any linguistic detection method to prove deception without confirmation by means external to the analysis. Research limitations/implications - The research was limited to English language studies in the linguistic detection of deception literature and to the re-calculation of data in the research literature. Whether the paper has implications for future studies depends on the success of two arguments that are made: the published research projects in the field reflect one or more of four methodological problems that create doubt about the validity and/or the practical utility of their results; and the linguistic detection of deception is subject to an epistemological problem which theoretically limits the ability of any linguistic method of detection to establish with certainty the status of any particular questioned statement. Originality/value - This is the first published paper to identify and discuss a possibly irresolvable epistemological issue in the detection of deception by linguistic means, as well as unresolved issues of methodology and of utility to law enforcement analysts that characterize the research and the detection models in this field. It is also the first published paper to deconstruct the simple hit rate (and its variants) in order to demonstrate its deficiencies.