There is limited evidence that preprogrammed feedforward adjustments, which are modified in people with neurological and musculoskeletal conditions, can be trained and whether this depends on the type of training. As previous findings demonstrate consistent delays in feedforward activation of the deep abdominal muscle, transversus abdominis (TrA), in people with recurrent low back pain (LBP), we investigated whether training involving voluntary muscle activation can change feedforward mechanisms, and whether this depends on the manner in which the muscle is trained. Twenty-two volunteers with recurrent LBP were randomly assigned to undertake either training of isolated voluntary activation of TrA or sit-up training to activate TrA in a non-isolated manner to identical amplitude. Subjects performed a trunk perturbation task involving arm movement prior to and after training, and surface and fine-wire electromyography (EMG) recordings were made from trunk and arm muscles. Following a single session of training of isolated voluntary activation of TrA, onset of TrA EMG was earlier during rapid arm flexion and extension, to more closely resemble the responses in pain-free individuals. The magnitude of change in TrA EMG onset was correlated with the quality of isolated training. In contrast, all of the abdominal muscles were recruited earlier during arm flexion after sit-up training, while onset of TrA EMG was further delayed during arm extension. The results provide evidence that training of isolated muscle activation leads to changes in feedforward postural strategies, and the magnitude of the effect is dependent on the type and quality of motor training.