Our ability in face recognition varies with familiarity of faces we encounter, and familiarity may affect both the representation and processing of faces. Here we proposed a model for classifying faces where familiarity varied along two continuous dimensions, frequency of contact and directness of contact, creating four possible types of familiarity. Hypotheses pertaining to recognizing faces of Types I (close friends), III (strangers), and IV (celebrity) familiarity were proposed and tested in Experiments 1A and 1B. Results showed effects of familiarity in that the higher the frequency of contact and the more direct the contact, the better face recognition (i.e., Type I > Type IV > Type III). Furthermore, recognition of Type-I faces was immune to variation in expression between the learned and test images, whereas recognition of Type-III faces was easily disrupted by change in expression. However, evidence for average face as the underlying representation for Type-IV faces was less clear than expected in that average faces derived from greater number of images failed to outperform those derived from fewer images. In Experiments 2 and 3, we explored the possibility of transforming familiarity type along the dimension of contact frequency by varying the numbers of exposure and expression of Type-III faces. Results showed that the original images were better recognized than those with altered expressions. Moreover, performance on faces with fewer exposures was impaired by variation in expression, especially when targets were old; however, with sufficient exposures, performance was enhanced by variation in expression when targets were new. We posited that variation in expression gives rise to generalization, whereas memory of the original image represents a consolidated trace of a specific expression leading to image matching. Taken together, the results lend support to the proposed familiarity model, which also provides a framework for examining factors that may help transform faces of one type of familiarity into another.

吾人臉孔辨識的能力會隨著對臉孔的熟悉度而改變，而熟悉度可能同時影響臉孔的內在表徵與處理。在本研究中我們提出一個臉孔分類的二維度模式，臉孔可以依據接觸的頻率以及接觸的直接程度區分成四種類型。實驗一A與一B檢驗辨識類型一(親近的朋友)、類型三(陌生人)以及類型四(名人)臉孔的假設。結果顯示接觸的頻率愈高並且接觸的方式愈直接，臉孔辨識的表現愈佳(亦即類型一優於類型四，而類型四優於類型三)。此外，類型一臉孔的辨識比較不會受到表情變化的影響，但是類型三臉孔的辨識非常容易因表情的改變而變差。不過，平均臉做為類型四名人臉孔內在表徵的支持證據卻不十分明確，因為藉由多數影像所建構之平均臉的表現，並未優於藉由少數影像所建構之平均臉。在實驗二及三中，我們操弄類型三臉孔出現的次數以及表情的變化，探討藉由接觸的頻率改變熟悉類型的可能性。結果發現在測試階段保持與學習階段同一影像的臉孔，其辨識上的表現較表情改變的臉孔為佳。此外，當出現的次數較少時，表情的變化會使辨識曾經出現過的目標臉孔表現較差；反之，當出現次數夠多時，辨識未曾出現過的目標臉孔會因表情的變化而表現較佳。我們認為表情變化導致臉孔辨識的類化效果；而對原來出現過的臉孔影像之記憶，則是代表對特定表情存留的痕跡，並因此導致較佳的辨識表現。總結來看，本研究的結果不但對所提出的臉孔熟悉度模式提供了支持的證據，並且指出哪些因素可以促成臉孔的熟悉性從一種型態變成另一種型態。