Sustained preferential processing of social threat cues: bias without competition?