Well over a year ago, Mark posted about the atomistic way some researchers use hi-tech to detect deception, including the MRI-scanner and voice stress analysis. There may be no end to the ways that law enforcement can misuse such information, especially when it is hawked by those who know only a little about it.
I leave the technological research on this topic to Mark and to others who know it better than I do. I'm more familiar with the equally atomistic, easy-fix, instructional programs that a few former police officers and others are selling to law enforcement agencies these days. These techniques can be just as susceptible to misuse as the hi-tech ones because they're oversimplified, atomistic, under-researched and can be outright misleading. Law enforcement officers face a constant problem of trying to figure out when a suspect is lying and they constantly hope to find ways to do this. A machine would be nice, but so would a less technological approach. Some of the currently available programs are called SCAN, the Reid Technique, and Statement Analysis, which I described briefly in an earlier post.
Clearly, it would be helpful if we could tell when the verbal and non-verbal behavior of people could give us the key to deceptive statements. Non-verbal signals got the first attention, loosely based on the very credible and promising research of Paul Ekman and his associates in San Francisco. One of his terms, "leakage," came into vogue very quickly, but it was far removed from Ekman's painstaking and complex research setting and it was translated in a grossly oversimplief manner by the commercial vendors of deception detection. They apparently relied on the theory that if Ekman's analysis of many simultaneous non-verbal clues found in many viewings of many videotapes of many subjects, including visual records from their heads to their feet, can offer some clues to deception, then one or two isolated clues by themselves, regardless of the suspect's culture, age, gender, or ethnicity, must enable the police to do the same thing, even in that emotional, instant setting of a police interrogation. For example, if suspects don't look the cop in the eye, they must be lying, despite the cultural basis that supports eye-aversion in known contexts.
The most common commercial program for spotting liars is based on the Statement Analysis technique. Suspects write an account of what happened on the day of the event in question. This approach contains no non-verbal clues, of course, but it's advocates claim that the following language features can help cops decide whether suspects are being untruthful:
Many questions can be asked about Statement Analysis. It has the advantage of getting a suspect's words on paper before the interrogator has a chance to foul things up. But it's based on little or no credible research that would indicate that these features are diagnostic and it's hard to know what is meant by "overly," "unexpectedly," "unusual," "marginally," "excessively," "unnecessary," "disproportionate," and "low." These features come most frequently from police officers' reported past experiences about what worked for them in the interviewing process.
I've heard this approach praised at academic meetings and I've read books and papers on it (are these overly detailed or unusual details?). I guess I'm being overly detailed here (am I spontaneously repeating marginally relevant details?). I wasn't impressed by the papers I've heard (am I displaying subjectivity from my external associations?). I may have forgotten to mention that I may not have been impressed by the speeches either (am I admitting memory loss, correcting spontaneously, and hedging a bit?). Have you noticed how many times I self-referenced here ... and paused ... and used unnecessary connectors ... and produced a disproportionate amount of language in my prologue? I'll let you perform a type-token analysis of my lexical diversity.
All I can say is that I must be lying.
Posted by Roger Shuy at August 1, 2007 01:31 PM