A rapid method for determining the carbonyl value of frying oils has been developed using Fourier-transform infrared (FTIR) spectroscopy and chemometrics. One hundred and fifty-six frying oils with different carbonyl values were collected from an actual potato frying process. FTIR spectra in the range of 4000-650 cm(-1) were scanned with a FTIR spectroscopy apparatus using the attenuated total reflectance (ATR) method. A good calibration model was obtained using the partial least-squares (PLS) regression method with full cross validation for predicting the carbonyl value of frying oils. For the model, the coefficients of determination (R(2)), standard errors of cross validation (SECV) and standard errors of prediction (SEP) were 0.99, 1.87 μmol g(-1) and 1.93 μmol g(-1), respectively. Moreover, standard deviation ratios of reference data in the validation sample set to the SEP were higher than 3. This study shows that the carbonyl value of frying oils can be successfully determined to a high accuracy using FTIR spectroscopy combined with PLS regression.