This paper describes the use of Fourier Transform Infrared (F..T.-I.R.) spectroscopy coupled with an Attenuated Total Reflectance (A.T.R.) accessory to quantify the individual sugars (glucose, maltose, maltodextrins...) in a mixture extracted from a process of wheat transformation into sugar. A.T.R. accessory allows to analyze very viscous media such as molasses by pouring the medium on a flat crystal. In the first step, model mixtures are prepared in the concentrations common to sugar industries and are analyzed by F.T.-I.R. spectrometer. Binary (glucose/maltose) and ternary (glucose/maltose/maltotriose) solutions are used. Spectra are recorded from 1 300 to 850 cm-1. Multivariate mathematical processing methods are applied to these spectra (Partial Least Squares and Principal Component Regression). The results show a very good discrimination between the different sugars : for instance in the binary mixture, the standard error of prediction for glucose (ranking form 100 to 200 g/kg) and maltose (ranking from 250 to 290 g/kg) are respectively equal to 1.66 g/kg and 1.07 g/kg. Then the same experiment is run on real samples added with known concentrations of glucose, maltose, maltotriose and maltodextrins. The major problem is the determination of maltotriose but glucose, maltose and maltodextrins are well analyzed.