The authors investigated the units of selective attention within working memory. In Experiment 1, a group of participants kept 1 count and 1 location in working memory and updated them repeatedly in random order. Another group of participants were instructed to achieve the same goal by memorizing the verbal and spatial information in an integrative way as a moving digit. The behavioral data showed that switching attention between properties of an integrated working-memory item was faster than switching between respective properties of different items. Experiment 2 demonstrated that this switching facilitation cannot be simply ascribed to the different amount of working-memory items maintained by the two groups of participants. Finally, by adopting a pure verbal task in Experiment 3, the authors observed the same binding facilitation, with the possibility of "location-based selection" excluded. They summarize the observations of all 3 experiments in the study and suggest both a location- and object-based mechanism for attention selection in working memory.