Java XML DOM: how are id Attributes special?

The javadoc for the Document class has the following note under getElementById.

Note: Attributes with the name "ID" or "id" are not of type ID unless so defined

So, I read an XHTML doc into the DOM (using Xerces 2.9.1).

The doc has a plain old <p id='fribble'> in it.

I call getElementById("fribble"), and it returns null.

I use XPath to get "//*[id='fribble']", and all is well.

So, the question is, what causes the DocumentBuilder to actually mark ID attributes as 'so defined?'