This study investigated the welfare consequences of training dogs in the field with manually operated electronic devices (e-collars). Following a preliminary study on 9 dogs, 63 pet dogs referred for recall related problems were assigned to one of three Groups: Treatment Group A were trained by industry approved trainers using e-collars; Control Group B trained by the same trainers but without use of e-collars; and Group C trained by members of the Association of Pet Dog Trainers, UK again without e-collar stimulation (n = 21 for each Group). Dogs received two 15 minute training sessions per day for 4-5 days. Training sessions were recorded on video for behavioural analysis. Saliva and urine were collected to assay for cortisol over the training period. During preliminary studies there were negative changes in dogs' behaviour on application of electric stimuli, and elevated cortisol post-stimulation. These dogs had generally experienced high intensity stimuli without pre-warning cues during training. In contrast, in the subsequent larger, controlled study, trainers used lower settings with a pre-warning function and behavioural responses were less marked. Nevertheless, Group A dogs spent significantly more time tense, yawned more often and engaged in less environmental interaction than Group C dogs. There was no difference in urinary corticosteroids between Groups. Salivary cortisol in Group A dogs was not significantly different from that in Group B or Group C, though Group C dogs showed higher measures than Group B throughout sampling. Following training 92% of owners reported improvements in their dog's referred behaviour, and there was no significant difference in reported efficacy across Groups. Owners of dogs trained using e-collars were less confident of applying the training approach demonstrated. These findings suggest that there is no consistent benefit to be gained from e-collar training but greater welfare concerns compared with positive reward based training.