A commercial electronic nose (e-nose) equipped with a metal oxide sensor array was trained to recognize volatile compounds emitted by potatoes experimentally infected with Ralstonia solanacearum or Clavibacter michiganensis subsp. sepedonicus, which are bacterial agents of potato brown and ring rot, respectively. Two sampling procedures for volatile compounds were tested on pooled tubers sealed in 0.5–1 L jars at room temperature (laboratory conditions): an enrichment unit containing different adsorbent materials (namely, Tenax s TA, Carbotrap, Tenax s GR, and Carboxen 569) directly coupled with the e-nose (active sampling) and a Radiello ™ cartridge (passive sampling) containing a generic Carbograph fiber. Tenax s TA resulted the most suitable adsorbent material for active sampling. Linear discriminant analysis (LDA) correctly classified 57.4 and 81.3% total samples as healthy or diseased, when using active and passive sampling, respectively. These results suggested the use of passive sampling to discriminate healthy from diseased tubers under intermediate and real scale conditions. 80 and 90% total samples were correctly classified by LDA under intermediate (100 tubers stored at 4 1C in net bag passively sampled) and real scale conditions (tubers stored at 4 1C in 1.25 t bags passively sampled). Principal component analysis (PCA) of sensorial analysis data under laboratory conditions highlighted a strict relationship between the disease severity and the responses of the e-nose sensors, whose sensitivity threshold was linked to the presence of at least one tuber per sample showing medium disease symptoms. At intermediate and real scale conditions, data distribution agreed with disease incidence (percentage of diseased tubers), owing to the low storage temperature and volatile compounds uncon-finement conditions adopted. & 2014 Elsevier B.V. All rights reserved.