This paper describes a partial least squares (PLS) based automatic procedure to correct for changes in the spectral contribution of the solvent or solvent mixtures from solute spectra recorded in these solvents. The procedure was developed for successful on-line Fourier transform infrared (FT-IR) detection in gradient high-performance liquid chromatography (HPLC) separations. It requires a reference FT-IR data set containing all possible combinations of the expected variation in solvent composition. Furthermore, a spectral region (A) in these spectra is required where the solvents show absorption but the analytes do not. This is the case for the system ACN:H(2)O, an often-applied solvent mixture in gradient HPLC, in the spectral region from 2300-2400 cm(-1). By focusing on (A) the developed numerical method selects an appropriate background spectrum from the reference data set, which is then subtracted from the analyte spectra. The method was programmed in Matlab and tested on different isocratic and gradient on-line reversed phase liquid chromatography-Fourier transform infrared spectrometry (LC-FT-IR) data sets. This work describes a successful method to perform eluent correction in on-line coupling of FT-IR spectrometry with gradient LC.