The goal of this study was to explore whether cortico-cortical interactions measured during the retention of visual information in working memory are sensitive to the alternation of the output modality, that is, to the way the stored information is going to be used. We designed an experiment in which participants were asked to memorize the spatially ordered arrays of letter-like shapes and, after a specified delay, reproduce them via either hand drawing, keyboard typing, or pronouncing aloud. During the delay period, the high-density electroencephalogram was recorded. The electroencephalogram (EEG) records were used to explore how different cortical areas interact while maintaining internal representation of perceived stimuli in working memory. We estimated effective connectivity in the θ frequency range for a preselected set of regions of interest via fitting vector autoregressive model to their activity in the source space. The general picture that emerges from the findings is that the output modality affects the way the visually presented information is encoded. While delayed copying relies more on the processing of pictorial codes, pronouncing task involves predominantly verbal encoding maintained via rehearsal, with typing task that showed a connectivity pattern similar both to the copying and pronouncing, seems to use both phonological and pictorial formats during the retention period.