Got your tin-foil on?
Okay, let's play.
Lets say most of your consumer goods are tagged with RFID. On your person, you have your shoes, your belt, your wallet, a watch, that sort of thing.
You walk through a device (knowingly or not) that has an RFID reader system in it. It records the GUID of each RFID chip and stores it as a single linked group in a database. Lets say this happens at a place where you have provided identification--such as a credit card at a store or gas pump.
Over time, you wear several pairs of shoes, maybe two wallets (one daily one dress), and a couple of watches. Once all those GUIDs are associated with you, then any time a statistically significant number of those GUIDs are tracked at any location that does not require ID, you can be identified. Say, the gates of a stadium, the door of a courthouse, whatever.
Every time you ID yourself at a location with an RFID reader (again, known or unknown), you add to or confirm the database about you. With a relatively small amount of data (RFID samples) a person or system could then uniquely identify you purely based on the combination of RFID numbers carried on your person.
Librarians who support RFID book checkout systems (because it does, in fact, provide a whole host of very, very handy new tools for librarians) claim that no one could tell from the RFID chip what book you have unless they have access to the database. Of course, if someone has access to the database, you don't need to read the RFID chip. BUT, someone (government agent or private individual) could read the GUID on the RFID chip of whatever book(s) are of interest to them, and then watch to see who walks out the door with that RFID tag tucked away in a bag after checkout. You wouldn't need access to the database to tell who had the book, only access to the output of the RFID reader at the door or another RFID reader elsewhere.
Class dismissed, you may remove your tinfoil now. :)