The soil organic carbon (SOC) associated with different soil fractions varies in the composition and dynamics. The present work is aimed to evaluate the potential of near infrared spectroscopy (NIRS) to predict SOC content in different soil fractions of black soils. SOC contents of 136 black soil samples in China were analyzed and the NIR spectra were collected using a VECTOR/22 (Fourier transform infrared spectroscopy). Partial least squares (PLS) regression with cross validation was used to develop calibrations between reference data and NIRS spectra (n=100) which were validated using an independent set of samples (n=36). Predictions for water-sieved aggregate associated organic carbon were generally good with R2 (coefficient of determination) ranging from 0.69 to 0.82 and the RPD (residual prediction deviation) from 1.2 to 1.8. NIRS well predicted the SOC in <53 μm mineral fraction (R2=0.97, RPD=5.4), but the prediction for SOC in 250～2 000 μm or in 53～250 μm particulate matter fractions was poor. However, the prediction for the SOC in 53～2 000 μm fraction was good (R2=0.79, RPD=2.2). In addition, NIRS very well predicted the SOC in fine particle fraction (<20 μm) (R2=0.93, RPD=3.8). Accordingly, NIRS showed a good potential to predict SOC in some soil fractions and could reduce tedious laboratory analysis.