Calibration and diagnosticity of confidence in eyewitness identification: Comments on what can be inferred from the low confidence-accuracy correlation

The relationship between eyewitness confidence and accuracy as measured by the psi point-biserial correlation has been described as poor or even nonexistent in the literature on lineup identifications. In this article, 3 arguments are made. First, a low point-biserial correlation is compatible with good or even perfect calibration (realism) of confidence, and the correlation provides no information about whether witnesses over- or underestimate the probability of a correct identification. Second, point-biserial correlations provide almost no information about whether confidence is diagnostic in the sense that it should be taken into account by the court when evaluating eyewitness identifications. Third, useful information is provided by calibration analysis and computation of diagnosticity indices. These arguments are illustrated with data from an experiment with photo-confrontations that rely on photo material used by the Swedish Police and where foils were selected by experienced police officers in the manner of routine investigations.