Back muscle forces contribute essentially to the whole-body vibration-induced spinal load. The electromyogram (EMG) can help to estimate these forces during whole-body vibration (WBV). Thirty-eight subjects were exposed to identical random low-frequency WBV (0·7, 1·0 and 1·4 m/s-2 r.m.s. weighted acceleration) at a relaxed, erect and bent forward postures. The acceleration of the seat and the force between the seat and the buttocks were measured. Six EMGs were derived from the right side of the m. trapezius pars descendens, m. ileocostalis lumborum pars thoracis, m. ileocostalis lumborum pars lumborum; m. longissimus thoracis pars thoracis, m. longissimus thoracis pars lumborum, and lumbar multifidus muscle. All data were filtered for anti-aliasing and sampled with 1000 Hz. Artefacts caused by the ECG in the EMG were identified and eliminated in the time domain using wavelets. The individually rectified and normalized EMGs were averaged across subjects. The EMGs without WBV exhibited characteristic patterns for the three postures examined. The coherence and transfer functions indicated characteristic myoelectric responses to random WBV with several effects of posture and WBV magnitude. A comprehensive set of transfer functions from the seat acceleration or the mean normalized input force to the mean processed EMG was presented.The results can be used for the development of more sophisticated models with a separate control of various back muscle groups. However, the EMG–force relationship under dynamic conditions needs to be examined in more detail before the results can be implemented. Since different reflex mechanisms depending on the frequency of WBV are linked with different types of active muscle fibres, various time delays between the EMG and muscle force may be necessary.