Researchers have made big strides in investigating and understanding resilient responding in the face of stress and adversity. Resilience is best understood as a process that unfolds over time in response to a stressor or potentially traumatic event (Bonanno Reference Bonanno2005; Bonanno & Diminich Reference Bonanno and Diminich2013). There is now accumulating evidence that resilience, defined as a trajectory of limited or no decreases in functioning over time, is a common response to such events. This approach entails an empirical characterisation of heterogeneous stress responses (Galatzer-Levy Reference Galatzer-Levy2014). Numerous studies have since adopted this approach through the use of modeling methods such as latent growth mixture modeling (LGMM), and they consistently have confirmed that resilient individuals compose the largest group following various types of stressors and traumas (e.g., Berntsen et al. Reference Berntsen, Johannessen, Thomsen, Bertelsen, Hoyle and Rubin2012; Bonanno et al. Reference Bonanno, Galea, Bucciarelli and Vlahov2006; Reference Bonanno, Mancini, Horton, Powell, LeardMann, Boyko, Wells, Hooper, Gackstetter and Smith2012; Galatzer-Levy & Bonanno Reference Galatzer-Levy and Bonanno2012; Reference Galatzer-Levy and Bonanno2014). These findings have led to a conceptual shift that Kalisch et al. take up in their present article, away from a focus on risk factors and psychopathology and towards a focus on protective factors and resilience.

The critical questions to be tackled, however, remain: Who adopts such a resilient response trajectory? What are the precise mechanisms leading to a resilient response? As empirical studies have indicated, multiple pathways to resilience exist, and these may sometimes be unexpected and achieved by means that might not be adaptive under other circumstances (Bonanno Reference Bonanno2005). Kalisch et al. identify several routes to a better understanding of such mechanisms and highlight the need to integrate findings from the neurosciences and animal learning.

Key neurobiological processes, often in interaction with environmental factors such as adverse experiences during childhood or trauma load, have been shown to affect psychological adjustment in the aftermath of exposure to trauma. A major challenge will be to integrate and translate such findings to clinical interventions and prevention efforts. We may, on the basis of neuroscience findings, be able to characterise individuals in the early aftermath of exposure to stress and adversity and to discriminate those who are likely to be resilient from those likely to succumb to stress (Galatzer-Levy et al. Reference Galatzer-Levy, Karstoft, Statnikov and Shalev2014a). Neurofeedback methods have been developed that may directly target patterns of brain functions that support resilience (Stoeckel et al. Reference Stoeckel, Garrison, Ghosh, Wighton, Hanlon, Gilman, Greer, Turk-Browne, deBettencourt, Scheinost, Craddock, Thompson, Calderon, Bauer, George, Breitner, Whitfield-Gabrieli, Gabrieli, LaConte, Hirshberg, Brewer, Hampson, Van der Kouwe, Mackey and Evans2014). Further strides will be necessary to pave the way for such efforts and to put research findings into practice by developing effective interventions that boost resilience.

Finally, human studies need conversely to inform preclinical animal learning studies. For example, it is increasingly understood that pathways to resilience and susceptibility to stress-related disorders are, at least in part, distinct in males versus females. Despite these findings and that stress-related disorders are significantly increased in women compared with men, the vast majority of animal models, however, have been conducted in male animals (Lebron-Milad & Milad Reference Lebron-Milad and Milad2012; Shanksy Reference Shanksy2015), which hampers transferability to both sexes in humans.

The task of identifying complex resilience mechanisms clearly benefits from a collaborative effort of researchers from different domains; that is, neuroscience, behavioral and cognitive science, and others. As pointed out by Kalisch et al., a common language and a shared conceptual framework will be critical to advancing the field. Resilience has been conceptualized many ways, but a straightforward and face-valid approach has been to identify a population that demonstrates positive adjustment in the face of adversity by disaggregating those individuals who demonstrate only transient stress or symptom responses following trauma, using methods such as LGMM (Bonanno & Diminich Reference Bonanno and Diminich2013). This same approach can be used to identify distinct trajectories of response in animal models of threat learning, extinction, and motivated behavior, thus increasing the translatability of such models to understand patterns of stress response including resilience, as animal models are key to the identification of neurobiological mechanisms (Galatzer-Levy et al. Reference Galatzer-Levy, Bonanno, Bush and LeDoux2013; Reference Galatzer-Levy, Moscarello, Blessing, Klein, Cain and LeDoux2014b). From molecules to circuits, to behavior and neuroendocrine response, to conscious and non-conscious cognitions and emotions, the responses to environmental threats and their aftermath unfold and interact in complex and dynamic ways over time, leading to sustained responses, or global organismic states (LeDoux Reference LeDoux2014). To add to the complexity, trauma exposure is not amenable to tight experimental control, therefore necessitating, at least in part, the use of naturalistic cohorts such as soldiers, natural disaster survivors, and individuals identified in emergency medical settings.

Given such complexity, the ability to identify causal mechanisms and develop predictive models for early identification may be hindered by traditional data analytic approaches and can benefit from recent advances in machine learning. These methods can integrate large sets of heterogeneous sources of information to predict, classify, and identify unique causal mechanisms leading to distinct trajectories of response (Aliferis et al. Reference Aliferis, Statnikov, Tsamardinos, Mani and Koutsoukos2010; Galatzer-Levy et al. Reference Galatzer-Levy, Karstoft, Statnikov and Shalev2014a). These approaches have potential to address the nagging limitations of traditional statistical approaches such as large variable-to-sample ratios, heterogeneous underlying distributions, the need for individual-level predictive accuracy, redundancy in data sources, and the need to discover relationships in the data that are not hypothesized a priori.

Such methodological innovations in the study of posttraumatic stress and resilience inevitably will shift the field from asking relatively limited questions regarding the direct effects of individual factors on resilient responding, to more-complex questions, such as when, how, why, and who is resilient. Ultimately, such an approach will improve the lives of those exposed to stress and adversity.