In ergonomics research, two-dimensional (2-D) biomechanical models are often used to study the mechanical loading of the low back in lifting movements. When lifting movements are asymmetric, errors of unknown size may be introduced in a 2-D analysis. In the current study, an estimation of these errors was made by comparing the outcome of a 2-D analysis to the results of a recently developed and validated 3-D model. Four subjects made two repetitions of five lifting movements, differing in the amount of asymmetry. The results showed a significant underestimation of the peak torque by 20, 36 and 61% when the initial position of a box was rotated 30, 60 and 90 degrees with respect to the sagittal plane of the subject. The main cause of this underestimation was a pelvic twist, resulting in an erroneous projection of a pelvic marker on to the sagittal plane due to pelvic twist. It is suggested that from 30 degrees box rotation a 2-D analysis may easily lead to wrong conclusions when it is used to study asymmetric lifting.