How and when to use causal and associational language

Deciding which concepts should be described in causal language and which should notResearch questions fall into one of three categories: descriptive, predictive, or causal. 1 The past decade has seen many calls for the appropriate use of causal language when the underlying research question is causal, 1-4 yet it remains common for researchers who are asking fundamentally causal questions to avoid causal language, particularly when articulating their research question. 5 6A recent random sample of articles from five of the top journals in epidemiology found that 69% appeared to have causal goals but used non-causal language (eg, association, risk factor, linked to, related to) in the research question. 5Here, we clarify why, if the underlying research question is causal, it is not only preferable to use causal language to articulate the question, but necessary.We also propose clear guidelines for how often misused terms such as "causal effect," "association," and "estimated causal effect" should be used in research articles and in reference to which quantities.Box 1 contains a short fictional dialogue demonstrating why causal research questions cannot be articulated using phrases such as "the association" or "an association" (or other non-causal language) because such language does not tell the reader which precise association is of interest.A research question stating that the goal is to "estimate the association," for example, could refer to a crude or adjusted association, and there are many possible adjusted associations that could be of interest.This phrasing is very common in medical research and epidemiology.Given the importance of the research question in conducting research, it is concerning that simple questions (eg, which association is the association) are not trivially easy to answer.If we cannot clearly explain what is meant by our research questions, how can we be expected to provide clear answers?