Tasks designed to deliberately expose the effects of an intervention by focusing on the aspects of the system that are being tested, minimising interaction with other elements. For example, if you have added a feature to help users keep track of what they are doing, a targeted task could include a lot of interruptions. Targeted tasks increase effect in the noise-effect-number triangle and so increase statistical power. However, targeting can be misleading if you are not careful, as the targeted task may not be representative of the full range of tasks that are likely to be carried out; that is generalisation can be difficult.
