The use of visible and near infrared spectroscopy (Vis/NIRS) was explored as a tool to detect the citric acid and tartaric acid in orange juice. Calibration models were developed from high performance liquid chromatography (HPLC) as a reference method and Vis/NIRS data using partial least squares (PLS) regression with cross-validation. '' Fingerprint '' was investigated based on PLS loadings and regression coefficients, offering theories for feature recognition and also providing possibilities to develop a simple and rapid device. The correlation coefficients (r) and root mean squares error of prediction (RMSEP) in the best model for citric acid was 0.944 and 0.596, and 0.930 and 0.013 for tartaric acid. The results indicate that it is very feasible to detect the quality of orange juice using Vis/NIRS technique. Comparing the parameters with those derived from HPLC determination of the extract obtained from the same samples, the results of Vis/NIRS analysis were not better than those of HPLC analysis with low coefficient of variation below 0.18% and high correlation coefficient over 0.9999. Nevertheless, due to the short consuming time, low cost of monitoring and a guaranteed high sample throughout, Vis/NIRS technique has its potential for the rapid, reliable and nondestructive detection of organic acids in orange juice. (C) 2007 Elsevier Ltd. All rights reserved.