It has been demonstrated that observation of grasping actions presented as point-light displays (PLDs) activates the action observation network (AON). The aim of this event-related fMRI study was to investigate the neural correlates of imitation of fully visible (FV) and PLDs manipulative actions within the AON. Twenty-six participants were instructed to imitate a hand rotating two objects in two directions (clockwise, anticlockwise) presented in two visual conditions: FV and PLDs. In each trial of each condition, participants had to observe, plan and imitate the presented action. Passive observation of the same FV and PLDs actions constituted the control conditions. Imitation was recorded using an MR-compatible camera and subsequently scored. The scoring revealed that motor performance was comparable in the two conditions, indicating that, despite conveying only kinematic information, also PLDs enable an accurate imitation performance as that driven by FV stimuli.Univariate fMRI analyses showed that during all events AON areas were similarly activated in both conditions. The contrast between FV and PLDs observation-to-imitate showed activations in visual areas, while the opposite contrast revealed bilateral activations of the insula, suggesting that FV-based imitation is mostly driven by pictorial information, while that based on PLDs mainly exploits kinematics. Using multivariate pattern analysis, we found that differences between FV and PLDs observation-to-imitate can also be decoded within parietal and premotor areas, suggesting a differential information decoding within the AON. These findings demonstrate that kinematics alone is sufficient to imitate an observed action, and that distinct stimuli features are decoded within the AON.