Processing of quantitative information, investigated with fMRI.

Ever since the discovery of the ‘number neurons’, the neural representation of quantity in the brain has been thought of as a number-selective coding system. In such a system, the neuron is activated by a specific quantity but numerically close quantities also activate the neuron. Recent fMRI studies also confirmed the existence of a number-selective system in humans. Several computational modelling studies predicted a number-sensitive coding stage as a necessary preceding stage to the number-selective neurons (Verguts & Fias, 2004). In this coding scheme, the coding is analogous to the number it represents. This can be implemented by neurons that respond monotonically to number (e.g., more strongly for larger numbers). Recently, the biological reality of such a system has been demonstrated by use of single-cell recording, in the lateral intraparietal area (LIP) of the macaque monkey. In this thesis, we searched for evidence of number-sensitive coding in humans. Using a priming paradigm, we found behavioural evidence for a number-sensitive system in humans for small non-symbolic numerosities (1 to 5). Using event-related fMRI, we showed number-sensitive activation in the human LIP area in the same number range. Remarkably, we could not extend these results for larger numerosities (2 to 64). Whereas the lack of results in the behavioural priming experiment could be due to an insensitivity of the method, this was not a plausible explanation in the fMRI experiment, as the activity measured in human LIP significantly decreased for numerosities larger than 8. We therefore concluded that the number-sensitive system is liable to a capacity limit for higher numerosities, which could be caused by the use of lateral inhibition. We further suggest that the implementation of this lateral inhibition is dependent on the particular task set, and that the capacity limit is not present (or less stringent) when numerosity is not behaviourally relevant. This could explain the finding of number-sensitive neurons for larger numerosities in monkeys. Finally, we suggest that a different mechanism is employed when numerical value of large numerosities is relevant. This leads to the conclusion that dot patterns in the small and large number range are processed differently