By investigating children whose parents have dyslexia, family risk (FR) studies are expanding our understanding of the intergenerational transmission of dyslexia. These studies, however, vary in their identification of FR, and how the use of different identification methods influences research findings and conclusions is yet to be systematically investigated. This study aims to evaluate the association between two FR identification methods--parental self-reports and direct skill assessments--and their unique contributions in the prediction of children's reading. The study employed two datasets: a prospective FR sample (half of the parents in the sample had dyslexia and the remaining half did not) and an unselected sample. Parental self-reports and direct skill assessments correlated strongly (0.60) in the prospective FR sample and moderately (0.42) in the unselected sample. Moreover, both FR identification methods were almost equally predictive of children's reading (explaining 5%-9% of the variance at different time points) in the prospective FR sample only. In the prediction of the children's skills, the two methods complemented each other only for some of the measures. At the same time, in the unselected sample, parental skills were not predictive of children's reading, whereas self-reports were. The two FR identification methods seem to have equally high predictive power when the variability in parental data is high. However, they lose their predictive power when either the lower or higher end of the parental reading distribution is underrepresented.