Soil organic carbon (SOC) models have been widely used to predict SOC change with changing environmental and management conditions, but the accuracy of the prediction is often open to question. Objectives were (i) to quantify the amounts of C derived from maize in soil particle size fractions and at various depths in a long-term field experiment using 13C/12C analysis, (ii) to model changes in the organic C, and (iii) to compare measured and modelled pools of C. Maize was cultivated for 24 years on a silty Luvisol which resulted in a stock of 1.9 kg maize-derived C m−2 (36% of the total organic C) in the Ap horizon. The storage of maize-derived C in particle size fractions of the Ap horizon decreased in the order clay (0.65 kg C m−2) > fine and medium silt (0.43) > coarse silt (0.33) > fine sand (0.13) > medium sand (0.12) > coarse sand (0.06) and the turnover times of C3-derived C ranged from 26 (fine sand) to 77 years (clay). The turnover times increased with increasing soil depth. We used the Rothamsted Carbon Model to model the C dynamics and tested two model approaches: model A did not have any adjustable parameters, but included the Falloon equation for the estimation of the amount of inert organic matter (IOM) and independent estimations of C inputs into the soil. The model predicted well the changes in C3-derived C with time but overestimated the changes in maize-derived C 1.6-fold. In model B, the amounts of IOM and C inputs were optimized to match the measured C3- and C4-derived SOC stocks after 24 years of continuous maize. This model described the experimental data well, but the modelled annual maize C inputs (0.41 kg C m−2 a−1) were less than the independently estimated total input of maize litter C (0.63 kg C m−2 a−1) and even less than the annual straw C incorporated into the soil (0.46 kg C m−2 a−1). These results indicated that the prediction of the Rothamsted Carbon Model with independent parameterization served only as an approximation for this site. The total amount of organic C associated with the fraction 0–63 μm agreed well with the sum of the pools ‘microbial biomass’, ‘humified-organic matter’ and IOM of the model B. However, the amount of maize-derived C in this fraction (3.4 g kg−1) agreed only satisfactorily with the sum of maize-derived C in the pools ‘microbial biomass’ and ‘humified organic matter’ (2.6 g kg−1).