In our daily life, we flexibly switch between unimanual and bimanual movements, e.g., we unimanually manipulate a smartphone and bimanually manipulate a tablet. However, how neural representations and behavioral aspects are related between those movements is still unclear. Here, by extending a framework of motor primitive (Thoroughman & Shadmehr, 2000, Nature), we propose a novel model of motor primitives that can unify motor learning in unimanual and bimanual reaching movements.

Our model based on neurophysiological results since activities of motor primitives likely correspond to neural activities. In reaching movements, activities of motor cortex neurons depend on movement directions and each neuron shows its maximal activity when the movement direction corresponds to its preferred direction (PD). Since neural activities are different between unimanual and bimanual movements (Rokni et al., 2003, JNs), we investigated the conditions in which a model of motor primitives can reproduce the following counter-intuitive results. Learning effects in bimanual movements towards "a fixed target direction" are " partially " transferred to unimanual movements (Nozaki et al., 2006, Nat Neurosci). Another study reported learning effects in bimanual movements towards "eight target directions" are " perfectly " transferred to unimanual movements (Wang et al., 2013, PLoS One). We analytically revealed that the PD modulation (difference of PD between unimanual and bimanual movements) is a necessary and sufficient condition to reconcile the partial and perfect transfer.

Secondly, we extend our model to reproduce other kinds of phenomena in motor learning in bimanual movements (Tcheang et al., 2009, EBR, Yokoi et al., 2011, JNs, Howard et al., 2010, Jnp). We revealed that the PD modulation is a necessary and sufficient condition to reproduce those kinds of phenomena.

In summary, we revealed that the PD modulation is a key factor to unify motor learning in unimanual and bimanual movements. Notably, the PD modulation was reported by neurophysiological experiments (Rokni et al., 2003, JNs). Furthermore, our theory leads to a testable prediction: the generalization function from bimanual to unimanual movements is smaller and wider than that function within bimanual movements.