In Central America, coffee (Coffea arabica) is commonly grown under different leguminous Inga tree species which are regularly pruned for shade control, and fuelwood production. Understanding tree-crop interactions in these systems, such as belowground competition for water and nutrients is limited by the difficulty to quantify the proportion of roots belonging to the different species. For example root morphologies of coffee and Inga densiflora are similar and root sorting by eye almost unfeasible. The aim of this study was to quantify the specific fine root mass distribution of crop and tree in a C. arabica - I. densiflora agroforestry system using near infrared spectrometry. The study was carried out in a 7 year old coffee plantation (4722 plants ha-1) shaded by I. densiflora (278 trees ha-1), established on a Costa Rican Andosol at 1200 m elevation under annual rainfall of about 2300 mm. The fine root (diam < 2 mm) biomass of mixed samples of coffee and Inga was investigated in soil cores (10 cm height, 8 cm diameter) down to 1m depth taken along the coffee rows at 50 cm from coffee plants and at 3 distances from the base of Inga trees. Pure samples of coffee and Inga roots (attached to the shoot) were also taken in the field. All samples were then oven-dried at 60°C and ground to a fine powder. Binary root mixtures containing known proportions (from 2.5 to 97.5%) of each plant species were artificially prepared by weighing and mixing root powders of the two species. Root samples of each species, artificial mixtures and all mixed root samples collected in the field were scanned with a NIRSystems6500. The spectral data were recorded at 10-nm intervals from 400 to 2500 nm. Coffee and tree species were spectrally different, particularly between 400 and 1300 nm. The calibration equations obtained by NIRS to estimate the proportion of species in root mixtures were accurate for the two species. These equations were consequently used to determine the specific mass r