In this study, a quantitative method for measuring the content of iota (iota), kappa (kappa), lambda (lambda), mu (mu) and nu (nu) carrageenan in powder samples from industrial production was developed using rapid spectroscopic techniques and chemometrics. The reference method for quantification was 600 MHz H-1 NMR spectroscopy, which was found superior to 150 MHz 13 C NMR due to higher sensitivity. Quantitative calibrations using partial least squares (PLS) regression on each carrageenan form were developed and compared for the three spectroscopic methods investigated, infrared, Raman and near-infrared spectroscopy. A large mixture design yielded near-perfect calibration models for all spectroscopic techniques, the best ones using the newly developed preprocessing method extended inverted signal correction (EISC) on Raman spectra yielding prediction errors (RMSECV) in the range 0.5-2.1% for the five different carrageenan forms. After variable selection using the new method backwards interval-PLS (bi-PLS), the models built on EISC treated Raman spectra of design samples as well as a few production samples yielded good predictions for all samples with RMSECV between 0.7 and 2.2% and correlations between 0.98 and 1.00. (C) 2004 Elsevier Ltd. All rights reserved.