Complex span tasks are predictive of many aspects of behavior, in both experimental and applied areas of cognitive psychology. Our view is that these tasks measure primarily working memory capacity (WMC), which we argue is the ability to control attention. The development of the Attention Network Test (ANT) provided the opportunity to study the relationship between WMC and specific types of attention. Extreme WMC-span groups differed in the executive control network but not in the alerting or orienting networks, supporting the view that individual differences in WMC reflect variation in the ability to control attention. We discuss problems with the design of the ANT that limit its appropriateness for applied research. Copyright © 2006 John Wiley & Sons, Ltd.