Partial directed coherence (PDC) as a frequency-domain representation of Granger casuality (GC) could detect both strength and direction of cortical interactions by multivariate autoregressive (MVAR) model of electroencephalography (EEG). In the present study, we investigate the underlying neural networks mechanisms of "rotational uncertainty effect" during mental rotation (MR) task by PDC analysis of multichannel EEG signals before and after the visual stimuli presented, we found that (i) temporally the "rotational uncertainty effect" involved an activated network before the visual stimuli presented, which could also affect the cognitive process of MR later; (ii) the causality functional connectivity network indicated that the bi-directional frontal [symbol see text] parietal networks played critical roles in maintaining the readiness during the MR task. These findings suggest that functional networks of un-cued preparation before visual stimuli presented are worth to be paid more attention. And these networks provide crucial casuality information to understand the neural mechanism for "rotational uncertainty effect" in MR task.