Measures of broad fluid abilities including verbal, quantitative, and figural reasoning are commonly used in the K-12 school context for a variety of purposes. However, differentiation of these domains is difficult for young children (grades K-2) who lack basic linguistic and mathematical literacy. This study examined the latent factor structure of a picture-based measure of broad fluid reasoning abilities using a bifactor MIRT model to separate general and broad domain factors in a large representative sample of U.S. school children. Substantial evidence showed that picture-based item formats can distinguish between general and domain-specific fluid reasoning abilities in the early school grades. The verbal tasks showed the strongest domain factor and discriminant validity, although the quantitative tasks also showed considerable evidence of a domain factor. Furthermore, comparisons of ELL vs. non-ELL, FRL-eligible vs. non-FRL, Black-White, Hispanic-White, and Asian-White students all yielded small to negligible group differences (below 0.4 "SD") on these measures. These results compare favorably to differences observed on tests using traditional item formats, and are smaller than the 0.50-1.0 "SD" group differences often observed in older students. (Contains 7 tables and 2 figures.)