Researchers have used eye-tracking measures to explore the relationship between face encoding and recognition, including the impact of ethnicity on this relationship. Previous studies offer a variety of conflicting conclusions. This confusion may stem from misestimation of the relationship between encoding and recognition. First, most previous models fail to account for the structure of eye-tracking data, potentially falling prey to Simpson's paradox. Second, previous models assume a linear relationship between attention (e.g., the number of fixations to a to-be-remembered face) and recognition accuracy. Two eye-tracking studies (Ns = 41, 59), one online experiment that manipulates exposure (N = 150), and a mega-analysis examine the effects of ethnicity using what we believe to be more appropriate analytical models. Across studies and measures, we document a novel, critical pattern: The relationship between attention and recognition is nonlinear and negatively accelerating. At low levels of baseline attention, a small increment in attention improves recognition. However, as attention increases further, increments yield smaller and smaller benefits. This finding parallels work in learning and memory. In models that allow for nonlinearity, we find evidence that central features (eyes, nose, and mouth) generally contribute to recognition accuracy, potentially resolving disagreements in the field. We also find that the effects of attention on recognition are similar for ingroup and outgroup faces, which have important implications for theories of perceptual expertise. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

中文翻译：

---

眼睛凝视与种族内群体和外群体面孔的识别准确性之间的非线性关系。


研究人员使用眼动追踪措施来探索人脸编码和识别之间的关系，包括种族对这种关系的影响。以前的研究提供了各种相互矛盾的结论。这种混淆可能源于对编码和识别之间关系的错误估计。首先，大多数以前的模型都无法解释眼动追踪数据的结构，可能会成为辛普森悖论的牺牲品。其次，以前的模型假设注意力（例如，对要记住的人脸的注视次数）和识别准确性之间存在线性关系。两项眼动追踪研究 （Ns = 41， 59），一项操纵曝光的在线实验 （N = 150） 和一项大型分析，使用我们认为更合适的分析模型来检查种族的影响。在研究和测量中，我们记录了一种新的关键模式：注意力和认可之间的关系是非线性的，并且呈负向加速。在基线注意力水平较低时，注意力的小幅增加会提高识别能力。然而，随着注意力的进一步增加，增量产生的好处越来越小。这一发现与学习和记忆方面的工作相似。在允许非线性的模型中，我们发现有证据表明，中心特征（眼睛、鼻子和嘴巴）通常有助于提高识别准确性，从而可能解决该领域的分歧。我们还发现，注意力对识别的影响对于内群体和外群体面孔相似，这对感知专业知识理论具有重要意义。（PsycInfo 数据库记录 （c） 2024 APA，保留所有权利）。