A prerequisite to L1 homophone effects in L2 spoken-word recognition