When two faces are not better than one: Serial Limited-Capacity Processing with Redundant-Target Faces

Many researchers believe that faces – whether presented alone or as part of an ensemble – are processed automatically. According to this idea: (a) the detection of single- or multiple- faces is resource-free and does not require allocation of attention, and (b) visual search for faces is held in parallel. The current study set to test these hypotheses directly. Participants performed in a redundant target detection task, responding according to the presence or absence of a face (or faces) on the display. We used a rigorous methodology known as the system factorial technology (SFT, Townsend & Nozawa, 1995), which afforded the simultaneous assessment of: (a) architecture (serial vs. parallel), (b) stopping rule (exhaustive vs. self-terminating), and (c) processing capacity (limited, unlimited, or supercapacity). SFT analyses on RT means and RT distributions pointed conclusively to a serial self-terminating architecture with limited-capacity. These findings cast serious doubts on the alleged automaticity of faces.

诸多研究者认为，无论人脸是单独呈现还是作为视觉集群的一部分，其加工过程均为自动化的。根据这一观点：(a) 单个或多个人脸的检测无需占用认知资源，亦无需分配注意力；(b) 人脸的视觉搜索以并行方式开展。本研究旨在直接检验上述两项假说。被试完成一项冗余目标检测任务，需根据屏幕上是否出现人脸（或多个人脸）做出对应反应。本研究采用了被称为系统析因技术（System Factorial Technology, SFT，Townsend & Nozawa, 1995）的严谨方法学，该技术可同时评估三类加工特征：(a) 加工架构（系列加工与并行加工）、(b) 终止规则（穷尽式终止与自我终止式终止），以及(c) 加工容量（有限容量、无限容量与超容量）。基于反应时均值与反应时分布的SFT分析明确指向了具备有限容量的系列自我终止加工架构。上述研究结果对所谓的人脸自动化加工假说提出了强烈质疑。