Near infrared sensor-based sorting is an emerging preconcentration technology which holds promise for many mineral processing applications, such as elimination of calcite and clay waste from ore. Preconcentration serves to increase the processing efficiency as well as to reduce the total processing cost through the rejection of unwanted gangue. Given small-scale heterogeneity in complex ores, i.e. assorted minerals occurring side-by-side inside an area of measurement, the total mineral composition may not be easily discerned from a near infrared spectrum. Hence, mineral identification and subsequent classification involves the analysis of absorption features in terms of feature depth, width, position, and level of spectral reflectance. This research investigates the near infrared spectral characteristics of minerals as a function of particle size fraction, specifically for individual minerals commonly found in malachite-rich copper ores. Samples of pure minerals are crushed and sieved into different size fractions and scanned with a near infrared line scanner. It was found that the presence of characteristic absorption features and their wavelength position were a better identification parameter than the reflectance level. The implication for preconcentrating a typical malachite-rich copper ore through near-infrared based sorting is discussed. (C) 2014 Elsevier Ltd. All rights reserved.