Numbers and mathematics are a substantial part of our culture and society considerably influencing our decisions on a daily base. Poor numeracy poses a serious burden for persons affected and makes it difficult to function effectively in our everyday’s life. The innate ability to process numerosities usually enables us to develop complex mathematical skills at a young age. However, children with developmental dyscalculia struggle with the acquisition of numerical and arithmetical skills. Despite the high prevalence of 3-7% and the importance of being numerate in society, development dyscalculia remains a neglected problem. Hence, little is known about the neuronal development of numerical cognition. The aim of the present thesis is to gain understanding about the development of numerical cognition in typically achieving children and, in particular, children with developmental dyscalculia. The intricate relationship of space and number was investigated considering the possibility of a generalised magnitude system. Longitudinal functional magnetic resonance imaging and behavioural data were acquired in order to trace typical and atypical developmental effects. Results of the longitudinal study reveal that in typical development numerical representations specialise with age and experience, resulting in a consistent and well-developed frontoparietal network. In contrast, persistent deficits in number processing and arithmetical skills are found in children with developmental dyscalculia compared to their peers. Brain imaging results suggest an age-related activation increase in number-specific regions pointing to a promising continuation in neuronal development in dyscalculic subjects. However, the increase in domain-general regions and the progression to a rather diffuse instead of a focussed number network corroborates the view of a delayed and inefficient developmental course. The results of the second study show that continuous and discrete magnitudes were both processed with high accuracy and almost identical neuronal networks. This favours the view that continuous and discrete magnitudes might rely on one generalised system for magnitude. Moreover, the ability to process continuous and discrete magnitudes is specialised in typically developing adolescents and seems to be preserved in subjects with developmental dyscalculia. Neuronal findings also point to the use of compensatory systems, suggesting a slight delay in the development of the discrete and continuous numerical system in developmental dyscalculia.The present data is only a first attempt to understand the typical and atypical developmental course of numerical cognition. Future studies need to put effort in the understanding of the complex dependencies and interconnections of the single factors in developmental dyscalculia leading to effective and practical implications for education and therapy.