Visuospatial working memory (VSWM) is a set of cognitive processes used to encode, maintain and manipulate spatial information. One important feature of VSWM is that it has a limited capacity such that only few items can be actively stored and manipulated simultaneously. Given the limited capacity, it is important to determine the conditions that affect memory performance as this will improve our understanding of the architecture and function of VSWM. Previous studies have shown that VSWM is disrupted when task-irrelevant eye movements are performed during the maintenance phase; however, relatively fewer studies examined the role of eye movements performed during the encoding phase. On one hand, performing eye movements during the encoding phase could result in a stronger memory trace because the memory formation is reinforced by the activation of the motor system. On the other hand, performing eye movements to each target could disrupt the configural processing of the spatial array because the spatial representation has to be updated with each movement to maintain perceptual stability. Therefore, this work was conducted to examine whether task-relevant saccadic eye movements performed during the encoding phase of a visuospatial working memory task affect the recall of serially presented targets. Results from two experiments showed that average recall accuracy was significantly higher when the spatial array (set size ≥ 7) was encoded using a covert strategy-that is, while participants fixated on a central target, in comparison to an overt strategy-that is, while participants moved their eyes to fixate on each target. Furthermore, the improvement in accuracy was evident only for targets presented in the first half of the sequence, suggesting that the primacy effect is modulated by the presence of eye movements. We propose that executing saccades during encoding could interfere with the ability to use a chunking strategy or disrupt active visualization of the configuration. In conclusion, this is the first study to show that task-relevant saccadic eye movements performed during encoding may actually reduce the spatial span of VSWM. These results extend the current knowledge about the role of eye movements in VSWM, and have implications for future studies investigating the VSWM.