Despite the importance of identifying individuals with reading disabilities, existing operational definitions of reading disability do not result in reliable identification. A large part of the problem arises from measurement error when a cut-point is imposed on a continuous distribution, especially for low base-rate conditions. One way to reduce measurement error is to include additional predictors in reading disability models. The present study examined co-occurring math disability as a possible additional criterion for predicting reading disability. Meta-analysis was used to examine the probability of individuals with reading disability also having a comorbid math disability. Possible moderators including age, severity of disability, and language were examined. The main result was an average weighted odds ratio of 2.12, 95% confidence interval [1.76, 2.55], indicating that students with a math disability are just over two times more likely to also have a reading disability than those without a math disability. Implications of the results are discussed.