臉孔辨識及其發展差異：多層次文獻回顧

Face Recognition and Its Developmental Differences: A Multi-Level Review

本文是一篇文獻回顧性的文章，全文主要分成兩大部分，第一部份是就臉孔辨識的議題做一廣泛且深入的介紹，從臉孔辨識異於一般物體辨識的「倒立效應」切入，並進而討論倒立效應之外的其他臉孔辨識的指標。接著，介紹臉孔辨識的一些特殊現象的相關研究，包括臉孔的半側臉效應、柴契爾效應、他族效應及負片效應等。然後進入臉孔辨識的大腦神經機制的研究，首先討論許多關於臉孔辨識的大腦神經生理研究，包括事件相關電位的與功能性核磁造影的，這些研究發現辨識臉孔時所引發腦部的反應異於辨識一般物體。接下來，討論不同類型的病人，這些病人共同的特點是他們辨識臉孔時都有不同程度的障礙；有些病人是因為腦傷引起的臉孔辨識障礙，有些則是基因造成的，甚至還有些病人是因為出生時受到視覺剝奪造成的。近年來，甚至發現有許多人雖無任何生理疾病或視覺剝奪的經驗，但是他們卻有先天性的臉孔失認症狀。之後，討論四類試圖訓練受試者辨識臉孔的研究，包括古力寶（Greebles）、倒立臉孔、跨種族臉孔，和發展訓練等；以及這些訓練研究及其發現對臉孔辨識機制的啟示。本文的第二部分將焦點轉移至臉孔辨識的發展議題，針對目前關於此議題的相關研究發現做較深入的探討與介紹，自嬰兒的臉孔偏好研究談起，接著將臉孔辨識的歷程區分成兩個切入的角度，包括「臉孔的整體處理與部件處理」及「臉孔的初階關係特性與二階關係特性」，並針對這兩個角度探討相關的辨識歷程與發展差異的研究發現。最後，針對臉孔辨識的研究現況，提出五個仍待進一步釐清或探討的問題，作為本篇多層次文獻回顧的結論。

This article extensively reviews the literature on mechanisms of face recognition and its developmental trajectories. In Part I, we begin with the inversion effect, an index widely considered as the hallmark of face recognition mechanism. We then discuss several other important effects of face perception, including three-quarter view, Thachterized faces, other-race effect, and the effect of negative images, along with their respective implications on face representations. Next, we review relevant brain imaging studies on normal population and prosopagnosic patients inflicted either by brain damage (acquired), genetic defects (developmental), or by early visual deprivation. Also, recent neurophysiological studies have also refueled the unresolved debate on whether faces are processed by specialized and dedicated neural mechanisms. At the end of Part I, we review four types of training studies: training on Greebles, inverted faces, other-race faces, and on various age groups, with particular emphasis on the limitations inherent in these training studies and the implications they yielded. Then in Part II, we shift focus to the developmental trajectories and age-related differences of face recognition performance. We review studies showing visual preferences for faces among newborns, followed by developmental differences shown in young children. The processes underlying face recognition are addressed by (a) holistic vs. component analysis, and (b) first- vs. second-order relational properties, and relevant studies that shed light along these two themes are covered. We conclude our multi-level review by listing five important questions that, in our view, await future efforts and further investigation.