Developmental prosopagnosia (DP) is a cognitive condition characterized by a relatively selective impairment in face recognition. Currently, people are screened for DP via a single attempt at objective face-processing tests, usually all presented on the same day. However, several variables probably influence performance on these tests irrespective of actual ability, and the influence of repeat administration is also unknown. Here, we assess, for the first known time, the test–retest reliability of the Cambridge Face Memory Test (CFMT)—the leading task used worldwide to diagnose DP. This value was found to fall just below psychometric standards, and single-case analyses revealed further inconsistencies in performance that were not driven by testing location (online or in-person), nor the time-lapse between attempts. Later administration of an alternative version of the CFMT (the CFMT-Aus) was also found to be valuable in confirming borderline cases. Finally, we found that performance on the first 48 trials of the CFMT was equally as sensitive as the full 72-item score, suggesting that the instrument may be shortened for testing efficiency. We consider the implications of these findings for existing diagnostic protocols, concluding that two independent tasks of unfamiliar face memory should be completed on separate days.

发育不全（DP）是一种以面部识别相对选择性受损为特征的认知疾病。当前，通过一次客观面部处理测试的尝试对人们进行DP筛查，通常都在同一天进行。但是，不管实际能力如何，几个变量可能会影响这些测试的性能，并且重复给药的影响也是未知的。在这里，我们首次评估了剑桥人脸记忆测试（CFMT）的重测可靠性，这是全球诊断DP的首要任务。发现该值刚好低于心理测验标准，单例分析表明，性能的进一步不一致不是由测试位置（在线或亲自）或两次尝试之间的间隔所驱动。后来发现使用CFMT的替代版本（CFMT-Aus）对确认边缘病例也很有价值。最终，我们发现CFMT的前48个试验的性能与72个项目的总评分一样敏感，这表明该仪器可能会缩短测试效率。我们考虑这些发现对现有诊断方案的影响，得出结论，两个不熟悉的面部记忆的独立任务应在不同的日期完成。