Scientific and mathematical thinking relies on the ability to evaluate whether conclusions drawn from conditional (if-then) arguments are logically valid. Yet, the neural development of this ability -- termed deductive reasoning -- is largely unknown. Here we aimed to identify the neural mechanisms that underlie the emergence of deductive reasoning with conditional rules in children. We further tested whether these mechanisms have their roots in the neural mechanisms involved in judging the likelihood of conclusions. In a functional Magnetic Resonance Imaging (fMRI) scanner, 8- to 13-year-olds were presented with causal conditional problems such as “If a baby is hungry then he will start crying; The baby is crying; Is the baby hungry?”. In Validity trials, children were asked to indicate whether the conclusion followed out of necessity from the premises. In Likelihood trials, they indicated the degree of likelihood of the conclusion. We found that children who made accurate judgments of logical validity (as compared to those who did not) exhibited enhanced activity in left and medial frontal regions. In contrast, differences in likelihood ratings between children were related to differences of activity in right frontal and bilateral parietal regions. There was no overlap between the brain regions underlying validity and likelihood judgments. Therefore, our results suggest that the ability to evaluate the logical validity of conditional arguments emerges from brain mechanisms that qualitatively differ from those involved in evaluating the likelihood of these arguments in children.