Default mode and dorsal attention network involvement in visually guided motor sequence learning

Motor sequence learning (MSL) paradigms are often used to investigate the neural processes underlying the acquisition of complex motor skills. Behavioral and neuroimaging studies have indicated an early stage in which spatial learning is prominent and a late stage of automatized performance after multiple training periods. Functional magnetic resonance imaging (fMRI) studies yielded both decreased and increased activations of the sensorimotor and association areas. However, task-negative and task-positive intrinsic connectivity networks (ICNs), the default mode (DMN) and dorsal attention (DAN) networks involved in governing attention demands during various task conditions were not specifically addressed in most studies. In the present fMRI study, a visually guided MSL (VMSL) task was used for bringing roles of visuospatial and motor attention into foreground in order to investigate the role of attention-related ICNs in MSL. Seventeen healthy, right-handed participants completed training and test sessions of VMSL during fMRI on the 1st day. Then, after daily training for three consecutive days outside the scanner, they were re-tested during the 5th day's scanning session. When test session after early learning period was compared with training session, activation decrease was observed in the occipito-temporal fusiform cortex, while task-related suppression of DMN was reduced. Reduced deactivation after early learning was correlated with decreased error rates. After late learning stage we observed activation decreases in bilateral superior parietal lobules of task-positive DAN, dorsal precunei, and cerebellum. Reduced activity in left posterior parietal and right cerebellar regions were correlated with gains in speed, error rate, respectively. This dissociation in activity changes of DMN and DAN related areas suggests that DAN shows high contribution during both early and late MSL stages, possibly due to attention requirement for automatization of spatial and temporal aspects of motor sequence. In contrast, spatial learning occurring during early MSL stage was sufficient for releasing DMN resources.