When evaluating items in a sequence, the current judgment is influenced by the previous item and decision. These sequential biases take the form of assimilation (shifting toward the previous item/decision) or contrast (shifting away). Previous research investigating facial attractiveness evaluations provides mixed results while using analytical techniques that fail to address the dependencies in the data or acknowledge that the images represent only a subset of the population. Here, we utilized cross-classified linear mixed-effects modeling across 5 experiments. We found compelling evidence of multicollinearity in our models, which may explain apparent contradictions in the literature. Our results demonstrated that the previous image's rating positively influenced current ratings, and this was also the case for the previous image's baseline value, although only when that image remained onscreen during the current trial. Further, we found no influence of the next face on current judgments when this was visible. In our final experiment, the response bias due to the previous trial remained present even when accounts involving motor effort were addressed. Taken together, these findings provide a clear framework in which to incorporate current and past results regarding the biases apparent in sequential judgments, along with an appropriate method for investigating these biases. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

中文翻译：

---

使用交叉分类模型对面部吸引力判断的序列效应：调查感知和反应偏差。


当评估序列中的项目时，当前的判断会受到前一个项目和决策的影响。这些顺序偏差采取同化（转向前一个项目/决定）或对比（转向远离）的形式。先前调查面部吸引力评估的研究提供了好坏参半的结果，同时使用的分析技术未能解决数据中的依赖性或承认图像仅代表人口的子集。在这里，我们在 5 个实验中使用了交叉分类的线性混合效应模型。我们在模型中发现了多重共线性的令人信服的证据，这可以解释文献中明显的矛盾。我们的结果表明，前一张图像的评级对当前评级有积极影响，前一张图像的基线值也是如此，尽管只有当该图像在当前试验期间保留在屏幕上时才如此。此外，我们发现，当下一张面孔可见时，下一张面孔对当前判断没有影响。在我们的最终实验中，即使涉及运动努力的账户得到解决，由于之前的试验而产生的反应偏差仍然存在。总而言之，这些发现提供了一个清晰的框架，其中纳入了有关连续判断中明显偏差的当前和过去的结果，以及调查这些偏差的适当方法。 （PsycInfo 数据库记录 (c) 2020 APA，保留所有权利）。