Rates of agreement among alternative definitions of reading disability and their 1- and 2-year stabilities were examined using a new measure of agreement, the affected-status agreement statistic. Participants were 288,114 first through third grade students. Reading measures were "Dynamic Indicators of Basic Early Literacy Skills" Oral Reading Fluency and Nonsense Word Fluency, and six levels of severity of poor reading were examined (25th, 20th, 15th, 10th, 5th, and 3rd percentile ranks). Four definitions were compared, including traditional unexpected low achievement and three response-to-intervention-based definitions: low achievement, low growth, and dual discrepancy. Rates of agreement were variable but only poor to moderate overall, with poorest agreement between unexpected low achievement and the other definitions. Longitudinal stability was poor, with poorest stability for the low growth definition. Implications for research and practice are discussed. (Contains 4 tables, 1 figure, and 1 note.)