Multi-level scope-chain tracing, closures that "remember" their outer scope after the outer function returns, and the classic lexical-scope trap where a function's scope is fixed by where it's written, not who calls it.