Context It is difficult to determine with accuracy the nutrition of bird diets through observation and analysis of dietary items. Collection of the ingested material from the birds provides an alternative but it is often limited by the small sizes of samples that can be obtained. Aims We tested the efficacy of near-infrared reflectance spectroscopy (NIRS) to assess the nutritional composition of very small samples of growing-parrot crop content. Methods We used 30 samples of the crop content of free-living scarlet macaw (Ara macao) chicks. Samples were scanned with a near-infrared reflectance analyser, and later analysed by traditional wet laboratory methods for crude protein/N, fat, ash, neutral detergent fibre, P, K, Ca, Mg, Cu, Zn and S. A calibration model was developed using principal components analysis. Key results Coefficients of determination in the calibration (R2) and standard errors of cross-validation (SECV) for most of the nutrients showed a good performance (mean R2 of 0.910.11s.d., n=10) when excluding Zn (R2 of 0.15, SECV=25.37). Conclusions The present results established NIRS as a valid technique for the non-destructive, low-cost prediction of a variety of nutritional attributes of avian crop contents as small as 0.5-g dry weight. Implications The use of NIRS expands the possibilities of wild-animal nutrition research.