A current issue in developmental science is that greater continuity in cognition between children and adults may exist than is usually appreciated in Piaget-like (stages or 'staircase') models. This phenomenon has been demonstrated at the behavioural level, but never at the brain level. Here we show with functional magnetic resonance imaging (fMRI), for the first time, that adult brains do not fully overcome the biases of childhood. More specifically, the aim of this fMRI study was to evaluate whether the perceptual bias that leads to incorrect performance during cognitive development in a Piaget-like task is still a bias in the adult brain and hence requires an executive network to overcome it. Here, we compared two numerical-judgment tasks, one being a Piaget-like task with number-length interference (called 'INT') and the other being a control task with number-length covariation ('COV'). We also used a colour-detection task to control for stimuli numerosity, spatial distribution, and frequency. Our behavioural results confirmed that INT remains a difficult task for young adults. Indeed, response times were significantly higher in INT than in COV. Moreover, we observed that only in INT did response times increase linearly as a function of the number of items. The fMRI results indicate that the brain network common to INT and COV shows a large rightward functional asymmetry, emphasizing the visuospatial nature of these two tasks. When INT was compared with COV, activations were found within a right frontal network, including the pre-supplementary motor area, the anterior cingulate cortex, and the middle frontal gyrus, which probably reflect detection of the number/length conflict and inhibition of the 'length-equals-number' response strategy. Finally, activations related to visuospatial and quantitative processing, enhanced or specifically recruited in the Piaget-like task, were found in bilateral posterior areas.