In the field of visual working memory (VWM) there exist a number of competing models that attempt to describe the capacity and storage features of memory. However, these models lack an explanation of how items are encoded. This presents a significant problem in the field, as it is currently difficult to distinguish between competing models such as the variable precision model (Van den Berg, Shin, Chou, George, & Ma, 2012) and the slots-plus-averaging model (Zhang and Luck, 2008). Given that these models have distinct theoretical accounts of memory, the lack of difference between the models’ behaviour creates a problem for their explanatory power. To create a point of difference in these theoretical accounts, we wanted to investigate the encoding state of VWM to see if how attention was distributed during a task. We used eye gaze as a proxy for attention to investigate connections between encoding and performance in VWM tasks. In standard tasks we found little connection between when and if the target was fixated and task performance. Instead, participants most frequently focus the centre of the screen and appear to learn to decrease their eye movements across the course of the experiment. We then switched to investigating a gaze contingent paradigm and saw not only a connection between fixation of the target and task performance, but most prominently an effect of recency. While some of our findings were more suggestive of the variable precision account compared to the slots-plus-averaging account, we conclude that the prominence of the recency effect is most in line with interference accounts of VWM (Oberauer & Lin, 2017). As well as our investigation into attention and performance, we also investigated task strategy. In our standard tasks, we saw a preference towards covert strategies that made few fixations however there were participants in these tasks that made consistent item fixations. This demonstrated individual differences in task approach. We found some consistent search strategies such as participants generally favouring an anticlockwise pattern when exploring study arrays that were presented on a circle. Given the individual differences we found, we recommend that strategic approach be investigated further to create a more complete understanding of VWM. In general, we caution the use of eye tracking in standard VWM tasks and believe that gaze-contingent designs should be utilised to explore the relationship between attention and VWM.