The Research Domain Criteria (RDoC) project's success rests on the assumption that constructs and data can be integrated across units of analysis and developmental stages. We adopted a psychoneurometric approach to establish biobehavioral liability models of sensitivity to social threat, a key component of potential threat that is particularly salient to the development of adolescent affective psychopathology. Models were derived from measures across four units of analysis in a community sample (n = 129) of 11- to 13-year-old girls oversampled for shy/fearful temperament. To test the ecological validity of derived factors, they were then related to real-world socioaffective processes in peer interactions over a 16-day ecological momentary assessment (EMA) protocol. Our results indicate that measures (i.e., amygdala reactivity to negative social feedback, eye-tracking bias toward social threat, parent- and adolescent-reports of social threat sensitivity) formed unit-specific factors, rather than one unified factor. These findings suggest that these factors were largely unrelated. Amygdala response to social punishment and attention bias toward threatening faces predicted real-world experiences with peers, suggesting that vigilance toward potentially threatening social information could be a mechanism through which vulnerable youth come to experience their peer interactions more negatively. We discuss measurement challenges confronting efforts to quantify developmentally sensitive RDoC constructs across units of analysis. (PsycInfo Database Record (c) 2022 APA, all rights reserved).