[ 29 ] E.K. Kemsley, J.K. Holland, M. Defernez, R.H. Wilson, J. Agric. Food Chem. 44 ( 1996 ) 3864. [ 30 ] O. Al-Jowder, E.K. Kemsley, R.H. Wilson, Food Chem. 59 ( 1997 ) 195. [ 31 ] H.M. Heise, R. Marbach, G. Janatsch, J.D. Kruse-Jarres, Anal. Chem. 61 ( 1989 ) 2009. [ 32 ] G. Janatsch, J.D. Kruse-Jarres, R. Marbarch, H.M. Heise, Anal. Chem. 61 ( 1989 ) 2016. [ 33 ] S.D. Stuchebryukov, V.M. Rudoy, Vibr. Spectrosc. 4 ( 1992 ) 95. [ 34 ] E.K. Kemsley, G.P. Appleton, R.H. Wilson, Spectro- chim. Acta 50A ( 1994 ) 1235. [ 35 ] R. Belali, J.-M. Vigoureux, J. Opt. Soc. Am. B 12 ( 1995 ) 2377. [ 36] F. Bartl, I. Delgadillo, A.N. Davis, J.P. Huvenne, M. Meurens, K. Volka, R.H. Wilson, Fresenius J. Anal. Chem. 354 ( 1996 ) 1. [ 37 ] S.R. Lowry, T.E. May, A. Bornstein, Appl. Spectrosc. 47 ( 1993 ) 30. [ 38 ] D.H. McQueen, R. Wilson, A. Kinnunen, Trends Anal. Chem. 14 ( 1995 ) 482. Reg Wilson is Head of the Food Composition and Authenticity Group at the Institute of Food Research, Norwich, UK. His research interests since 1984 revolve around the use of infrared spectroscopy for the analysis of foods. This includes on-line analysis, authentication, biopolymer functionality and the use of microspec- troscopic techniques for the study of the architecture of plant cell walls. Henri Tapp completed his PhD at Swansea where he investigated the application of a large horizontal self- resonant helical wave-guide for estimating body composition. Research interests at the Institute of Food Research, Norwich ( 1995^1998 ) included electrical capacitance tomography, and mid- and near-infrared spectroscopy. He is currently working at Camborne School of Mines, Redruth, on the application of sensor fusion techniques for monitoring £owing particulate systems. Kohonen arti¢cial neural networks as a tool for wavelength selection in multicomponent spectro£uorimetric PLS modelling: application to phenol, o-cresol, m-cresol and p-cresol mixtures R. Todeschini*, D. Galvagni Milano Chemometric Research Group, Department of Environmental Sciences, University of Milan, I-10126 Milan, Italy J.L. V| èlchez, M. del Olmo, N. Navas Department of Analytical Chemistry, University of Granada, E-18071 Granada, Spain Different strategies for wavelength selection for partial least squared (PLS) calibration models have been proposed. In this article, Kohonen arti¢cial neural networks ( K-ANN ) are used to select optimal sets of wavelengths for PLS calibration of mixtures with stray overlapping. This kind of variable selection appears simple and very effective due to the well known high correlation of spectroscopic data; a measure of the multivariate correlation of the different wavelength subsets is also given. This strategy has been applied to the resolution of mixtures of phenol, o-cresol, m-cresol and p-cresol by spectro£uorimetry. The number of samples to obtain the calibra- tion matrix is also reduced with respect to the number necessary when the full spectrum is used, and the predictive ability of the PLS method is improved. z1999 Elsevier Sci- ence B.V. All rights reserved. Keywords: Kohonen arti¢cial neural network; Multivariate calibration; Variable selection; Multivariate correlation; Phenol; Cresol; Excitation £uorescence 0165-9936/99/$ ^ see front matter ß 1999 Elsevier Science B.V. All rights reserved. PII:S0165-9936(98)00097-1 *Corresponding author. trends in analytical chemistry, vol. 18, no. 2, 1999 93