A quantitative assay for guar gum in carob gum, based on the extraction of proteins in acetonitrile–water (7:3), separation by capillary electrophoresis and multiple linear regression (MLR) using the areas of nine selected peaks as predictors, was improved by performing the extraction in the presence of gamanase. In the absence of the enzyme, peak migration times and areas depended on the guar content, which complicated peak identification and evaluation. Manual correction of the migration times by comparison with standard electropherograms obtained with pure carob and carob–guar mixtures was required; however, when the proteins were extracted under sonication at 60 °C for 30 min in the presence of gamanase, the migration times and peak areas did not vary with the composition of the carob–guar mixtures, and its reproducibility improved largely. These effects were attributed to the reduction of the viscosity of the extracts and the removal of the galactomannose interactions with the proteins and the capillary walls. Peak identification and evaluation were easily and directly performed on these electropherograms without further processing. An MLR model constructed with 36 carob–guar mixtures containing up to 20% guar, and by measuring the areas of 12 selected peaks and using eight of them as predictors, yielded a detection limit of 2.8% guar (α=β=0.05 criterion). A model of similar quality was obtained by partial least-squares (PLS) regression.