The item that matched the content of working memory (WM) could affect the orientation of visual attention, namely visual attention turned to choose the space or objects that were maintained in WM. According to the theory of optical complementary color, when the two colors of complementary pairs appear at the same time, it can cause the high contrast, and forms the largest bright in color vision. So in present study, according to the theory of contrast effect of complementary color, we adopt the experimental paradigm that combines the memory task with the visual search task, and introduce the background color to highlight the color of memory item and search target, thus it influences the visual selective attention that is based on the content of WM. In present study, we designed the two single factor bi-level within-subject experiments to separately operate the background color on the color of memory item and search target, in order to investigate whether there is a strong automaticity on the content of WM guide visual attention. In memory task, the participants were required to memorize the color of the memory sample. After that, the visual search display was presented. Search sequences consisted of six colored squares with a gap on their sides, the participants had to indicate which side of the target had a gap quickly and accurately. At the end of the trial, a single colored square was presented as a memory test probe. The participants had to judge whether or not the color of the probe was identical to the color of the memory sample. In experiment 1, the color of backgrounds and memory items were complementary color from beginning to end in order to highlight the memory items. In experiment 2, the color of backgrounds and target were complementary color all along to highlight the target. The results were as follows: (1) On the condition of operating the colors between background and memory items, the effect of the contents of WM was significant, and the RTs were significantly longer in the match condition than in the mismatch condition, t(1,11)=3.027, p<0.05. This finding supported that the memory-matching object had a privileged status to capture attention, even when there was absolutely no benefit for subjects to bias attention in favor of the memory match, and this capture had a strong automaticity. (2) On the condition of operating the colors between background and target, the effect of the contents of WM was not significant, and the RTs between the match condition and the mismatch condition did not reach the significant level, t(1,11)=1.76, p=0.106. We suggested that when target had a salience in vision scenes, the items that were maintained in WM could not capture visual attention. Therefore, the item that matched the content of WM could direct visual attention automatically, but that direction had certain conditions.