A substantial body of research has demonstrated that human and nonhuman animals have perceptually-based abilities to process magnitudes of nonsymbolic ratios (e.g., ratios composed by juxtaposing two-line segments). In prior work, we have extended the neuronal recycling hypothesis to include neurocognitive architectures for nonsymbolic ratio processing, proposing that these systems might support symbolic fractions acquisition. We tested two key propositions: (1) children should show neural sensitivity to nonsymbolic fractions before receiving formal fractions instruction, and (2) they should leverage this foundation by recruiting neural architectures for nonsymbolic fractions processing for symbolic fractions. We compared nonsymbolic and symbolic fractions processing among 2nd-graders (n = 28, ages 7.5-8.8), who had not yet received formal symbolic fractions instruction, and 5th-graders (n = 33, ages 10.3-11.9), who had. During fMRI scanning, children performed ratio comparison tasks, determining which of two nonsymbolic or symbolic fractions was larger. Both cohorts showed behavioral and neural evidence of processing nonsymbolic and symbolic fractions magnitudes, with performance modulated by numerical distance between stimuli. Consistent with our predictions, 2nd-graders recruited a right parietal-frontal network for nonsymbolic fractions but not for symbolic fractions, whereas 5th-graders recruited a bilateral parietal-frontal network for both, overlapping with but extending beyond that of 2nd-graders. Furthermore, nonsymbolic-symbolic neural similarity in the intraparietal sulcus was greater for 5th-graders than for 2nd-graders. These results present the first developmental neuroimaging evidence that neural substrates for nonsymbolic ratios exist before formal learning, which may be recycled to process symbolic fractions.