Researchers commonly use the Implicit Association Test (IAT) to assess the automatic attitudes of individuals and groups. Although contended by some, the IAT is used in large part due to its psychometric properties, which are generally superior relative to most other measures of automatic cognition. Much focus has therefore been dedicated to the IAT's psychometric properties (particularly its internal consistency). However, this work has focused near-exclusively on moderators based on the procedural features of the IAT itself, and little on the varying properties of the construct under investigation within the measure. This is despite the fact that attitude features have already been demonstrated to influence explicit attitude measures. Here, we intend to investigate whether different attitude features can effectively predict the internal consistency of IAT scores using a large-scale IAT dataset (lowest N = 30,161, highest N = 30,502). We find that five of six attitude features (personal importance, degree of thinking, certainty, self-concept, and most strongly polarity) are positively related to the reliability of the IAT. Our findings have significant implications for the way in which the IAT's reliability has been conceived.

研究人员通常使用内隐关联测试 (IAT) 来评估个人和群体的自动态度。尽管有人提出异议，但 IAT 的使用在很大程度上是由于其心理测量特性，相对于大多数其他自动认知测量方法而言，这些特性通常更优越。因此，很多焦点都集中在 IAT 的心理测量特性（尤其是其内部一致性）上。然而，这项工作几乎完全集中在基于 IAT 本身程序特征的调节器上，而很少关注该措施中正在调查的结构的不同属性。尽管态度特征已经被证明会影响外显态度测量。在这里，我们打算使用大规模 IAT 数据集（最低N  = 30,161，最高N  = 30,502）研究不同的态度特征是否可以有效地预测 IAT 分数的内部一致性。我们发现六个态度特征中的五个（个人重要性、思考程度、确定性、自我概念和最强烈的极性）与 IAT 的可靠性呈正相关。我们的研究结果对 IAT 可靠性的构思方式具有重要意义。