The purposes of this study were to examine the dimensions underlying morphological awareness (MA) in Arabic (construct validity) and to determine how well MA predicted reading (predictive validity). Ten MA tasks varying in key dimensions (oral vs. written, single word vs. sentence contexts, and standard vs. local dialect) and two reading tasks (real word and pseudoword reading) were administered to 102 Arabic-speaking Grade 3 children in Abu-Dhabi. Factor analysis of the MA tasks yielded one predominant factor, supporting the construct validity of MA in Arabic. Closer inspection revealed that this factor had two subcomponents, oral and written. Hierarchical regression analyses, controlling for age and gender, indicated that both the one- and the two-factor solutions accounted for 48% of the variance in word reading, and 40% of the variance in pseudoword reading, supporting the predictive validity of MA. Implications for future research, assessment, and instruction are discussed.