To reveal the functional organization underlying human cognitive and behavioral flexibility at resting stateWe hypothesized that functionally flexible regions are heterogeneous, supporting multiple cognitive components and temporally integrating information from functionally specialized brain networks in accordance with task demands. Moreover, a broad body of experimental work has demonstrated the dynamical organization of resting-state brain activity.1 We therefore proposed that the dynamics of intrinsic functional connectivity are expressed as a probabilistic configuration for each brain region. Two metrics were utilized to quantify this probabilistic configuration: (1) a complexity measure (entropy) reflecting functional flexibility (heterogeneity), and cross-module probability (CMP) measuring functional integration. This probabilistic model (Fig.1) includes the following steps: 1. We adopted the commonly used sliding window approach2 to estimate the dynamic functional connectivity matrices. 2. For a given brain region i, we reserved its 5 strongest functional connections at each sliding window. Thus, a normalized probability distribution p(Cij) could be obtained for the i’s dynamic functional connectivity using the following formula: $$p(C_{ij})=\frac{n(C_{ij})}{m\times nw},j=1,2,...N, and j\neq i$$ where n(Cij) is the number of connections between i and j across all time windows; m represents the number of connections for each sliding window (m=5); and nw denotes the number of sliding windows. 3. A complexity measure Hi was defined based on the probability distribution of region i using Shannon entropy: $$H_{i}=-\sum_{j=1}^Np(C_{ij})\times \log_{2}{p(C_{ij})}$$ 4. We assessed CMP by summing the probability of connections between i and regions in the other modules. Before calculating CMP, we divided the brain network into 6 modules based on traditional modular analysis.3 Thus, the CMP of region i was calculated using the formula: $$CMP_{i}=\sum_{j=1}^Np(C_{ij})\mid M(i)\neq M(j),j=1,2,...N, andj\neq i$$ where p(Cij) is the probability distribution for i’s dynamic functional connectivity; M(·) denotes the module that region belongs to. In order to remove the effect of modular size on CMP, we calculated a corrected CMPi as follows: $$CMP_{i}\mid corrected=\frac{CMP_{i}}{1+(N-n(i))/N}$$ where N is number of nodes of the whole brain network; and n(i) denotes the size of the module that node i belongs to. This framework was applied to resting-state functional MRI scans (TR=2000ms, TE=30ms, 220 volumes for each scan) acquired from 33 right-handed healthy volunteers (age, mean±sd=34.7±7.8 years) to investigate the functional representations of hand preference. Scanning was performed on a Siemens Trio 3.0 Tesla MRI scanner (Siemens, Erlangen, Germany). Informed consent was obtained from all subjects or their guardians. This study was approved by the Institutional Ethics Committee of East China Normal University (Shanghai, China)We found the left primary motor cortex dominant to handedness most frequently exhibited functional connectivity with regions in the sensorimotor network and left frontoparietal control network. In contrast, the right primary motor cortex dominant to non-handedness most frequently exhibited functional connectivity only with regions in the sensorimotor network (Fig.2). Furthermore, a paired t test revealed that both functional flexibility (p<0.001) and integration (p<0.01) of the left primary motor cortex dominant to handedness were significantly higher than those of the right primary motor cortex dominant to non-handedness (Fig.3).Hand preference denotes the individual predisposition to consistently use the right or left hand for most types of skilled movements. Handedness has been attributed to hemispheric asymmetry between bilateral cortical association areas controlling the cognitive-motor requirements of skilled movements.4 In particular, frontal and parietal circuits are intimately involved in the control of goal-directed movements.5, 6 Therefore, the heterogeneous and integrated function of the contralateral primary motor cortex may contribute to the skilled movement of the preferred hand.