Relational match-to-sample is a difficult task for young children. However, it has been shown that either presenting two examples of the relation or adding a label to a single presentation can improve children’s performance. The role of labels has been seen as increasing the likelihood of comparing the instances available. In this paper, we present sustained attention as an alternative to this view. Children completed a relational match-to-sample task in different conditions while an eyetracker registered their eye movements. When only one instance was available, children benefited from the addition of a label. This benefit was associated with an overall decrease in switching behavior, indicating greater sustained attention. Moreover, in the absence of a label, children who showed greater sustained attention were able to achieve good performance by the end of the task.