Abstract
Letter-name identification has been widely used as part of early screening to identify children who might be at risk for future word reading difficulty. The goal of the present study was to examine whether a reduced set of letters could have similar diagnostic accuracy rather than a full set (i.e., 26 letters) when used as a screen. First, we examined whether a hierarchical scale existed among letters by using a Mokken scale analysis. Then, we contrasted diagnostic accuracy among the 5, 10, 15, and 20 easiest letters, with all 26 letters by using receiver operating characteristic (ROC) curves and indices of sensitivity, specificity, positive predictive power, and negative predictive power. Results demonstrated that a hierarchical scale existed among items in the letter-name knowledge test. In addition, assessing students on the easiest 15 letters was not statistically distinguished from all 26 letters in diagnostic accuracy. The implications of the results for the use of a Mokken scale analysis in educational research are discussed.