A response to Pabst and Maurage: In defense of empathy self-report measures.

We agree with Pabst and Maurage [1] that improving the construct validity of empathy questionnaires will help advance the quality of our science. However, despite the shortcomings noted by Pabst and Maurage [1], cognitive and affective empathy, as assessed by the questionnaires in our meta-analysis [2], have distinctive heritability [3] and differentially predict many relevant psychological and health outcomes beyond alcohol use disorder (AUD) [e.g. 4, 5]. The empathy questionnaires included in our meta-analysis are also widely used by alcohol researchers [6] and by leading researchers in clinical [3, 7], health [8, 9] and social psychology [10, 11]. Given their predictive validity and wide use across the psychological sciences, it is important to quantitatively synthesize them in relation to AUD, particularly because there is broad interest in looking at the empathy-AUD association. We disagree that empathy behavioral tasks should be prioritized over self-report questionnaires for ‘indexing objective ability.’ Pabst and Maurage argue that behavioral tasks are preferred because individuals with AUD may be poor judges of their empathic abilities and self-report empathy measures correlate weakly with behavioral measures. They suggest that empathy behavioral tasks accurately convey information about empathy, whereas self-reports do not. We find this problematic for two reasons. First, there are theoretical and methodological explanations for weak associations between empathy self-report and behavioral measures beyond assuming a lack of insight about empathic abilities. Theoretically, weak associations may be because empathy self-report and behavioral tasks require different response processes [12-14]. Behavioral tasks focus on in-the-moment ‘behavioral snapshots’ of empathic abilities and may be influenced by contextual features (e.g. mood, motivation), whereas self-report questionnaires ask individuals to reflect on and average their empathy across various real-life situations. Methodologically, weak associations between empathy self-reports and behavioral measures may be because of the latter’s poor psychometric properties and/or lack of ecological validity. Empathy behavioral measures often have poor internal consistency and only weakly associate with other social cognition behavioral tasks [14]. The low reliability of empathy behavioral measures directly limits correlations that can be observed between them and empathy self-reports [13]. Empathy behavioral measures also lack ecological validity [15, 16]; they do not require participants to actually interact with people, but rather to make inferences based on pictures/videos/vignettes of people. Assessing empathy in real-time social encounters is needed [17]. Second, despite weak correlations between empathy self-report and behavioral measures, both may still predict relevant behavioral outcomes, as is the case with impulsivity [18]. Indeed, when we used a battery of impulsivity self-report and behavioral tasks in a large sample of participants, it was the impulsivity questionnaires that predicted externalizing behaviors with medium to large associations, whereas associations for behavioral tasks with externalizing outcomes were non-existent or small [19]. In summary, we do not share Pabst and Maurage’s [1] enthusiasm for empathy behavioral tasks, at least not for the ones currently used in the field. Self-report measures of empathy play a fundamental role in the young science of empathy, and a meta-analysis can help reveal places where there may be important predictive associations, based on theory. Like Pabst and Maurage [1], we believe more attention needs to be paid to the relevance of measurement issues in social cognition addiction research. We discuss these issues, offer an organizing theoretical framework, and provide suggestions for future work in this area in a chapter of an edited series on new directions in addiction science [16]. We are grateful for the opportunity to further discuss these important issues and commend Pabst and Maurage [1] for pointing out key issues in the science of social cognition and addiction. Kasey Creswell: Conceptualization; funding acquisition; writing—original draft. Lakshmi Kumar: Investigation; writing—review and editing. This paper was supported by grant R01 AA025936 to K. Creswell. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. None.