Implications of stress-pattern differences in spoken-word recognition