A weak object-based theory considers visual working memory to be made up of many subsystems. Rather than competing for memory resources, this theory suggests that information about different feature dimensions is stored in independent subsystems that each has a limited capacity of memory resource.In addition to describing the storage capacity limitations of different feature dimensions, supporters of weak object-based theory also argue that the binding between features can itself be a dimension of information to be stored. There remains controversy concerning whether the storage of binding information is processed automatically and whether it needs attentional resources. Treisman et al. (2002) suggested that binding in visual memory might require attention to be focused on maintaining the links between features during the delay. Similarly feature-integration theory proposes that, when multiple objects are present, focused attention is required to correctly bind features for initial perception. In contrast, Allen et al. (2006) by adding a digit-span task, explored whether feature binding required additional resources. They found that memory for bound conditions did not require more attention than memory for single feature conditions, and suggested that binding in itself did not require attentional resources. To examine whether or not binding needs attentional resources we measured participants’ ERPs in two conditions. In the single feature condition participants were required to judge whether the color of objects had changed from the initial display. On change trials, the objects occupied the same position but one item had changed to a new color that had not appeared in the initial display.In the feature binding condition, on no-change trials, participants had to judge both whether the color and the location in the test display had changed from the initial display. For change trials, the colors of any two objects exchanged position so that the relationship between colorand location changed for two objects. We compared the amplitude differences between the two conditions in the Contralateral Delay Activity (CDA) component of ERP data, to examine whether feature binding required attention. Accuracy for each condition was greater than 75% and there wereno significant behavioral differences between the single feature condition and the binding condition.Importantly, the ERP resultsalso showedno significantdifferences inamplitude across the two conditions. There was a significant main effect of set size. There were equal amplitudes for 3 items and 4 items, but amplitudes were significantly larger for both 3 and 4 item than for 2 items. At the same time, there was no significant interaction between condition and number of items. All of the results support the hypothesis that no additional memory resources are required to store the binding between two features (color and position of an item) compared to those required to store a single feature (color). These results confirm that the binding relationship needs no attention resources. We take the result to be evidence of a relatively automatic visual feature binding mechanism in working memory.