Madole & Harden (M&H) propose a discrimination between knowledge of causality and, on the contrary, correlation in behavior genetic studies, with a particular emphasis on genome-wide association studies (GWASs). Far from being a dichotomous difference, we argue that the boundary between these two types of causal relationships is indeed continuous, and any distinction would be arbitrary. Correlation itself is based on a counterfactual assumption, that is, that there is no relationship between the studied factors (Benesty, Chen, Huang, & Cohen, Reference Benesty, Chen, Huang, Cohen, Cohen, Huang, Chen and Benesty2009). The line between counterfactuality and causality seems to be briefly addressed in the manuscript, but only implied in both the introduction and in the rest of the article. We agree with the original authors that counterfactuality is not the only acceptable framework of causality, as counterfactuality itself might not prove to be either specific or sensitive to causality (Baumgartner, Reference Baumgartner2008). In fact, recent developments in the regularity theory of causation (based on the premises that causes are regularly followed by their effects; Baumgartner, Reference Baumgartner2008) have allowed for a precise estimation and quantification of confidence in causal relationships (Baumgartner, Reference Baumgartner2008), while also offering the opportunity to assess causality beyond dichotomous categories (whether “ill” or “healthy,” or expressing a certain behavior or not). These developments allow researchers to move toward approaches encompassing the possibility of evaluating spectra of continuum in traits, possibly through fuzzy-logic algorithms (Baumgartner & Ambühl, Reference Baumgartner and Ambühl2020), granting a more flexible evaluation of the interactions between such traits and specific genes. Specific research questions and study designs might then benefit from a careful operationalization of causality in investigation protocols, defining outcome variables either dichotomous, when methodologically sensible, or continuous, in the majority of cases where complex traits need to be considered.

M&H also argue that genetic analyses might not be causally informative, as the relationship between genotype and phenotype can be complex in behavior research. A critique is moved about the results of genetic studies, which are in some cases misquoted or overinflated. Indeed, no single study can inform our understanding of the complex biology behind the interaction of genes and traits. There is a real and actual risk of adopting “evidence-based” policies on frail epistemological and scientific grounds. This risk, however, is contemplated by most contemporary researchers and most of the general public (Visscher, Brown, McCarthy, & Yang, Reference Visscher, Brown, McCarthy and Yang2012). Nonetheless, supporting a careful review of evidence before adopting interventions does not invalidate the scientific knowledge gained by conducting behavior genetic studies. Even if GWASs were only to offer “association” knowledge rather than “true” causal understanding of underlying factors, it is relevant to point out that other means are available in genetic research to reach this goal, which may be more oriented toward “deeper” and “mechanistic” causal analysis (e.g., pathway analysis, protein functionality studies, endophenotype studies; Kendler & Neale, Reference Kendler and Neale2010); but in order for these tools to be used effectively, because of their high relative costs and technical complexity, candidate genes must be identified previously through GWASs. In fact, GWASs have offered considerable insight into nearly every field of contemporary biological sciences, from pharmacodynamics and pharmacokinetics, to proteomics and transcriptomics (Visscher et al., Reference Visscher, Brown, McCarthy and Yang2012).

A clarification of the term “causality” may also aid in delimiting the space of the discussion. Any operative definition of “causality” should consider scientific endeavors in an integrative manner, and as the result of multidisciplinary efforts. In fact, “science” can be defined as a continuous dialectical pursuit (Popper, Reference Popper1940), where “knowledge” is constantly updated with evidence derived from different sources. Therefore, scrutiny over preliminary evidence is indeed warranted before it is allowed to inform clinical or policy interventions. However, a primary goal is also to seek a balance between addressing the neglected needs of an individual and violating the right not to be harmed, which is a statutory principle that has guided the field of medicine since its inception. For these reasons, the risk of invalidating exploratory evidence in neuroscience in general, and behavioral genetics in particular, needs to be discussed. The hazard to implicitly propose new criteria in research, that is to consider “association studies” as secondary or even detrimental, should be critically evaluated, as it might severely impact both patients and researchers. For instance, limited funding and scarcity of resources may favor those capable of conducting large-scale “mechanistic” studies, harming scientific independence, tilting the balance in favor of consortium-led enterprises, with negative consequences on originality, scrutiny, and productivity in research (Wang, Veugelers, & Stephan, Reference Wang, Veugelers and Stephan2017). Large-scale “mechanistic” studies may also worsen the over-representation of white Anglo-Saxon, European, or East-Asian individuals in genetic studies (Sirugo, Williams, & Tishkoff, Reference Sirugo, Williams and Tishkoff2019). Additionally, research on several clinical conditions may never reach the volume necessary to conduct a large-scale investigation (Rosenberg & Finn, Reference Rosenberg and Finn2022), and no existing knowledge at present may properly guide causal “mechanistic” studies. Especially in those fields of medicine interested by complex behaviors (e.g., psychology, neurology, psychiatry), low prevalence and clinical heterogeneity burden the ease-of-access to interventional programs, as well as the inclusion in observational studies (Mitchell, Maki, Adson, Ruskin, & Crow, Reference Mitchell, Maki, Adson, Ruskin and Crow1997). However, it is possible to adopt mitigating options. For example, longitudinal designs can reach higher statistical power than cross-sectional ones, increasing replicability and confidence in the association between genes and phenotype (Rosenberg & Finn, Reference Rosenberg and Finn2022). Again, GWASs offer a cost-effective opportunity to first assess associations between genes and traits in these populations, and later inform more targeted protocols or interventions. Finally, the same conditions interested by low prevalence or high heterogeneity demand urgency in describing causal relationships, as they are taxed by a high rate of inadequacy in treatment (Bulik, Reference Bulik2021). For these reasons, invalidating behavioral genetic studies solely on concerns of describing causal associations may severely impact those individuals who they may benefit the most.