Because the local variables of one method can't be accessed by another method -- this is an architectural restriction that can't be broken. So when an anymous inner class object is created, what happens "behind the scenes" is that the anonymous object has some member variables that are initialized to have the same values as the local variables the inner class is using. The inner class code really uses these member variables. If the locals were not final, then you could tell that this is what was happening -- i.e., you could change the locals without the anonymous class noticing, or vice-versa. This restriction is there, then, basically to maintain the illusion that the anonymous class is accessing the local variables of the containing method.
Excerpt form SYBEX Java 2 Certification: �Any variable, either local variable or a formal parameter can be accessed by methods within an inner class provided that variable is marked final. A final variable is effectively a constant so this might seem to be a quite severe restriction, but the point is simply this: An object created inside a method is likely to outlive the method invocation. Because local variables and method arguments are conventionally destroyed when their method exists, these variables would be invalid for access by inner class methods after the enclosing method exits. By allowing access only to final variables, it becomes possible to copy the values of those variables into the object itself, thereby extending their lifetimes.� >R