Abstract
All studies are inherently biased, but some are more biased than others. This variation on a key theme from George Orwell's Animal Farm underscores a significant issue in public health. Ultimately, optimizing public health begins with understanding population health-particularly when assessing the impact of specific health risks that are often intertwined with both benign and malign health determinants. The objective of this contribution is to provide an overview of sources of bias in epidemiological research, drawing inspiration from the work of Rudolph Agricola-Northern Europe's first humanist and a homo universalis. Agricola's methodological approach distinguished between different categories of informational sources, which he deliberately employed as instruments for structured argumentation. This article presents a contemporary variation of that approach in the form of a complementary taxonomy, outlining examples of material and procedural bias sources that, individually or in combination, can affect estimates of mental health problems. These include the nature of the outcome itself and the context of the sample-covering its vulnerability and exposure profile, as well as broader population characteristics-along with data collection methods and analytical techniques. The value of this structured approach to disentangling bias in modern population health research is illustrated with examples from recent studies on the impacts of disasters and the COVID-19 pandemic. Researchers are encouraged to be modest, to carefully consider "locations" or "origins" of bias, and to interpret study findings with caution-especially when using them to inform public health policy or to make arguments about the nature and severity of population health issues.