In this study, the authors examined the validity of a holistically scored retell within a confirmatory factor analysis framework by comparing the fit of a three-factor model of reading with the data from a diverse sample of seventh and eighth graders. The final model demonstrated adequate fit, [chi][squared](32) = 97.316; comparative fit index = 0.96; Tucker-Lewis index = 0.94; and root mean square error of approximation = 0.08. Retell's chi-square difference, [delta][chi][squared](1) = 16.652, p less than 0.001, and factor loading (0.250, p less than 0.001) were higher for the comprehension construct. Similarly, retell's correlation to comprehension measures (r = 0.155-0.257, p less than 0.01) was stronger than its relationship to measures of fluency (r = 0.158-0.183, p less than 0.01) or word identification (r = 0.132, p less than 0.05). However, retell had a large residual variance (0.938) and low interrater reliability ([kappa] = 0.37), suggesting that improvements to the instrument are needed. Despite overall latent differences, retell did not demonstrate differential item functioning. (Contains 8 tables and 2 figures.)