Objective The ability of working memory (WM) is to temporarily hold and manipulate information for cognitive tasks.The aim of this study was to investigate the causal flow distribution of multi-channel electroencephalographs (EEGs) during Sternberg visual WM task to help understand WM mechanism from the view of neural information transform.Methods Sixteen-channel EEGs were recorded from 8 healthy subjects during Sternberg task,total EEG data for 80 trials were recorded and preprocessed.The EEG powers for each channel were calculated via Fast Fourier transform (FFT).The feature channel was extracted based on the maximum power,and the feature frequency band for WM was determined by time-frequency analysis.The directed transfer function (DTF) and the value of causal flow in EEG feature frequency band were calculated.Results The EEG power focused on θ band (4-8 Hz) and at Fz channel.The maximum value of causal flow was located at Fz,and the first 4 maximum negative values of causal flow were at C3,C4,C5,C6.Conclusions The feature frequency band for WM is the θ band.The causal source (responsible region) is at Fz (frontal region) and the causal sinks are at C3,C4,C5,C6.