The language of accurate and inaccurate eyewitnesses

Experimental psychologists have – for decades – espoused the unreliability of eyewitness identifications, but the advent of new statistical techniques such as confidence-accuracy characteristic analysis has revealed that eyewitness identifications are much more reliable than previously thought. When an eyewitness identifies the suspect with high confidence from an initial and properly-administered lineup, for example, that suspect is highly likely to be the person who originally committed the crime. The way confidence is collected in the laboratory – using a numeric rating scale – differs from the way confidence is collected in the real world – often by asking eyewitnesses to express their confidence in their own words. What is the best method for collecting an eyewitness’s level of confidence? To answer this question, we applied a novel machine-learning methodology to investigate the natural language of accurate and inaccurate eyewitnesses. This method revealed that verbal confidence statements provide much diagnostic information about the accuracy of identifications. Moreover, verbal confidence statements provide unique diagnostic information that is not otherwise captured by traditional indicators of identification accuracy such as numeric confidence ratings. However, the diagnostic value of a verbal confidence statement depends in part on the face recognition ability of the eyewitness: the natural language of strong face recognizers is more diagnostic than the natural language of weak face recognizers. These results are theoretically interesting, but from an applied perspective, this machine-learning methodology may prove useful to those in the criminal justice system that must evaluate eyewitnesses’ verbal confidence statements.