The fermentability of lignocellulose hydrolyzates have been predicted from the responses of a combination of chemical gas sensors. The hydrolyzates were prepared by dilute-acid hydrolysis of wood from pine, aspen, birch, and spruce. The volatile emission from the hydrolyzates before fermentation was measured, and the sensor array response pattern was compared with the observed fermentability of the hydrolyzates, i.e. with the final ethanol concentration after fermentation and the maximum specific ethanol production rate. Two concentration parameters in the hydrolyzates, furfural and the sum of furfural and 5-(hydroxymethyl)furfural (HMF), were also predicted from the responses. The sensors used were metal oxide semiconductor field effect transistors (MOSFET), tin oxide semiconductor devices, and conductive polymer sensors configured in two sensor arrays. The sensor array response pattern was analyzed by principal component analysis and artificial neural networks. Predictions from artificial neural networks deviated from measured values with less than 15%.The fermentability of lignocellulose hydrolyzates have been predicted from the responses of a combination of chemical gas sensors. The hydrolyzates were prepared by dilute-acid hydrolysis of wood from pine, aspen, birch, and spruce. The volatile emission from the hydrolyzates before fermentation was measured, and the sensor array response pattern was compared with the observed fermentability of the hydrolyzates, i.e. with the final ethanol concentration after fermentation and the maximum specific ethanol production rate. Two concentration parameters in the hydrolyzates, furfural and the sum of furfural and 5-(hydroxymethyl)furfural (HMF), were also predicted from the responses. The sensors used were metal oxide semiconductor field effect transistors (MOSFET), tin oxide semiconductor devices, and conductive polymer sensors configured in two sensor arrays. The sensor array response pattern was analyzed by principal component analysis and artificial neural networks. Predictions from artificial neural networks deviated from measured values with less than 15%.