The measurement of subtle beat-to-beat morphologic variability is complicated by the presence of respiration-induced movements of the heart. A statistical signal model is developed which accounts for scaling, rotation and time synchronization of a VCG loop. The maximum likelihood (ML) estimator of the model parameters is derived and tested on a material of high-resolution VCG recordings. The results show that the new method, relying on a spatial relationship of the leads, is able to efficiently reduce the effects of respiration on beat-to-beat variability.