Complex movements require the interplay of local activation and interareal communication of sensorimotor brain regions. This is reflected in a decrease of task-related spectral power over the sensorimotor cortices and an increase in functional connectivity predominantly in the upper alpha band in the electroencephalogram (EEG). In the present study, directionality of information flow was investigated using EEG recordings to gain better understanding about the network architecture underlying the performance of complex sequential finger movements. This was assessed by means of Granger causality-derived directed transfer function (DTF). As DTF measures the influence one signal exerts on another based on a time lag between them, it allows implications to be drawn on causal relationships. To reveal causal connections between brain regions that are specifically modulated by task complexity, we contrasted the performance of right-handed sequential finger movements of different complexities (simple, scale, complex) that were either pre-learned (memorized) or novel instructed. A complexity-dependent increase in information flow from mesial frontocentral to the left motor cortex and, less pronounced, also to the right motor cortex specifically in the upper alpha range was found. Effective coupling during sequences of high complexity was larger for memorized sequences compared with novel sequences (P = 0.0037). These findings further support the role of mesial frontocentral areas in directing the primary motor cortex in the process of orchestrating complex movements and in particular learned sequences.