JLS7 11.2.3 makes an analysis of catch blocks: "It is a compile-time error if a catch clause can catch (§11.2) checked exception class E1 and a preceding catch block of the immediately enclosing try statement can catch E1 or a superclass of E1."

There is no reason to restrict the analysis to checked exception classes. The desired principle is that the first catch of exception class E preempts later catches of E and subclasses of E. (Because that's the runtime behavior.) The word "checked" should be changed to "an".

Note that the preceding 11.2.3 rule - where a catch of a checked exception class is compared to a try block - is OK. It's not possible to prove an error if all you have is one catch of an _unchecked_ exception class and a try block; you must look at _other catch clauses_ to prove an error (namely, redundancy due to preemption).

Finally, 11.2 should say "We say that a statement or expression can throw ***an exception*** class E..." rather than "a checked exception class E". Whether a class is a checked exception class or an unchecked exception class is irrelevant for some rules in 11.2.3, and those rules that care will restrict themselves to checked exception classes.