Chasing consistency: On the measurement error in self-reported affect in experiments

How feelings change over time is a central topic in emotion research. To study these affective fluctuations, researchers often ask participants to repeatedly indicate how they feel on a self-report rating scale. Despite widespread recognition that this kind of data is subject to measurement error, the extent of this error remains an open question. Complementing many daily-life studies, this study aimed to investigate this question in an experimental setting. In such a setting, multiple trials follow each other at a fast pace, forcing experimenters to use a limited number of questions to measure affect during each trial. A total of 1398 participants completed a probabilistic reward task in which they were unknowingly presented with the same string of outcomes multiple times throughout the study. This allowed us to assess the test–retest consistency of their affective responses to the rating scales under investigation. We then compared these consistencies across different types of rating scales in hopes of finding out whether a given type of scale led to a greater consistency of affective measurements. Overall, we found moderate to good consistency of the affective measurements. Surprisingly, however, we found no differences in consistency across rating scales, which suggests that the specific rating scale that is used does not influence the measurement consistency.