If a language is being widely used in specific existing 'legacy' projects and for people learning it to get a job in their maintenance (e.g. COBOL?), then I'd say it's dead. If during the last five years every organization considering what language to use for their new production system (as opposed to someone's toy project or learning project) chose something else, then that language apparently has no valid niche remaining and is effectively dead, being chosen only due to legacy/compatibility reasons if the system was already started with it.
As the parent poster said - PHP4 sites exist in production and people work on PHP4 code maintaining them, and PHP4 is undeniably dead despite all that.
Also, I'm not sure whether I'd equate Clojure with Lisp - I'd treat it as a separate language/ecosystem evolved from Lisp as it was commonly understood; I'd consider Clojure activity as evidence that "Lisp may be dead but its descendents live on".