From the Kaspersky Lab Weblog:
When we upload something embarrassing about ourselves to, let's say Facebook, that's completely our fault. But there are other subtle ways to get information about us. Let's say a few words about tracking.
Every time you visit a website you request HTML that will be rendered in your local browser. This code may include external references, so you will request them as well. Nothing to be afraid of so far. [Screenshot]
But what happens when these external requests are used to track you? Is that possible? Let's suppose you are enjoying your favorite social network, let's say Facebook again. Even after you logout, your browser stores some cookies that identify you in Facebook. Then you visit some random website (www.randomwebsite.com) that includes any "I like" buttons, which in fact are external references to Facebook. And as Facebook has access to its own cookies, it gets a request with HTTP-Referer www.randomwebsite.com and with the cookie that identifies you. Result: Facebook knows all the sites you browse as far as they have a reference to Facebook.
Continued : http://www.securelist.com/en/blog/208193399/Where_is_my_privacy