Measuring the Diagnostic Performance of a Lineup

Recently, the argument has been made that receiver operating characteristic (ROC) analysis should be used to compare the diagnostic performance of different lineup procedures. However, a drawback to ROC analysis is that is requires multiple correct and false ID rates for each lineup procedure being compared. When only a single pair of correct and false ID rates is measured, what dependent measure should be used? Here, we contrast the use of d' with the diagnosticity ratio using the data reported by Carlson and Carlson (in press) and other previously reported data. Despite being based on a theory that was developed for list memory procedures, we show that, in practice, d' performs surprisingly well for lineup procedures. Moreover, d' far outperforms the diagnosticity ratio. We recommend that d' – not the diagnosticity ratio – be used as a dependent measure to compare the diagnostic performance of different lineup procedures.