Tree Testing is a task based research method whereby you supply a "tree", much like a sitemap, taxonomy or information architecture, and ask people to complete common tasks such as "Where would you find the latest mobile phones?". Once you've got some data to look at you'll be able to see which labels in your structure are working and which ones are the misleading "evil attractors" in order to improve the findability of your content.
It is important to validate your information architecture early in a project because it informs much of your navigation design and is a crucial part of your content strategy. After all, without great content that people can actually find, why would people want to visit your website? Validating that people understand the information architecture that you're using before you begin visual and interaction design will save you time in the long run.
People often wonder why Treejack doesn't include an ability to put navigation items into a footer or sidebar (like PlainFrame, for example). Although in principle this seems like a good idea, by doing this you cannot really know whether the information architecture you've chosen will stand alone, no matter what aesthetic design or which helper links the user interface designer chooses to use or what graphic changes are made over time. Wireframe testing is a useful tool but should come after Tree Testing as your layout and visual element start coming into play. By testing these things in succession you can confidently identify which elements of your overall design require change as a result of user testing.