Distinct extractable organic matter (EOM) fractions have been used to assess the capacity of soils to supply nitrogen (N). However, substantial uncertainty exists on their role in the N cycle and their functional dependency on soil properties. We therefore examined the variation in mineralizable N and its relationship with EOM fractions, soil physical and chemical properties across 98 agricultural soils with contrasting inherent properties and management histories. Mineralizable N was determined by aerobic incubation at 20 °C and optimum moisture content for 20 weeks. We used multivariate statistical modelling to account for multi-collinearity, an issue generally overlooked in studies evaluating the predictive value of EOM fractions. Mineralization of N was primarily related to the size of OM pools and fractions present; they explained 78% of the variation in mineralizable N whereas other soil variables could explain maximally 8%. Both total and extractable OM expressed the same soil characteristic from a mineralization perspective; they were positively related to mineralizable N and explained a similar percentage of the variation in mineralizable N. Inclusion of mineralizable N in fertilizer recommendation systems should be based on at least one OM variable. The most appropriate EOM fraction can only be identified when the underlying mechanisms are known; regression techniques are not suitable for this purpose. Combination of single EOM fractions is not likely to improve the prediction of mineralizable N due to high multi-collinearity. Inclusion of texture-related soil variables or variables reflecting soil organic matter quality may be neglected due to their limited power to improve the prediction of mineralizable N.