Sir,As far as I could make out, in response to the Learning Curve article[1] on the importance of inadequately measured, unmeasured, and unknown confounds, Verma[2] makes four broad points: (1) that the study of Coburn et al.[3] was methodologically adequate; (2) that my criticism of the study[3] was unjustified; (3) that the authors[3] did not intend to imply a cause–effect relationship between antenatal depression and infant health; and (4) that my article[1] did not cite references to support the arguments for the existence of potential confounds.Learning Curve articles are intended to teach young academicians about statistics, research methods, how to read a research paper, and related matters of a scientific and academic nature. Such teaching is best understood with the help of examples. As clearly indicated in the abstract, the observational study of Coburn et al.[3] was presented as an example, and the message that the teaching conveys applies not only to this study[3] but to all studies with an observational research design.With regard to the first point, observational studies can never ever be methodologically adequate for the very reason that my article[1] describes: it is difficult to identify and satisfactorily measure all the known confounds and impossible to measure unknown confounds. As a simple example, there is a long list of reasons why babies may experience common health disturbances during the first 12 weeks of life; just how many of these did Coburn et al.[3] record and adjust for in their analyses? If prospective studies cannot be comprehensive, then retrospective studies are even less likely to be comprehensive in their identification and measurement of confounds, and in adjustment for confounds.This does not mean that observational studies should not be performed. Such studies are necessary for hypothesis generation. However, such studies can also be found dreadfully wrong at the time of hypothesis testing,[4] and so a prudent reader would want to know whether or not to take the findings of an observational study at face value or to read between the lines and consider alternate explanations that include confounding.[5]With regard to the second point, I am sure that Coburn et al.[3] did their best with the available resources. However, responsible execution of a study does not make an observational study a gold standard in research, as I have explained earlier.With regard to the third point, Coburn et al.[3] indeed did imply causality; in fact, they went so far as to use the word “teratogenic” in the abstract of their paper, in the context of the potential effect of antenatal depression on fetal and infant health outcomes.With regard to the fourth point, Learning Curve articles, similar to other articles in the journal, have word count and reference restrictions. In any case, one does not reference the obvious. For example, it has been known for decades that maternal nutrition during pregnancy can affect fetal and infant health. Moreover, does one need to reference a statement that poor access to quality medical care can be responsible for poor maternal and infant health? Finally, unknown confounds cannot be listed, numbered, or even referenced precisely because they are unknown in current scientific knowledge, and it would take a courageous scientist to deny their existence. In this context, given the known impact of genetic predispositions on a wide range of biological outcomes, it would be foolish to discount their existence as potential confounders.Diligent readers of journals would be well aware that authors adjust for dozens, and sometimes well over 50 confounding variables when examining associations in observational studies. The take-home message here is that, outside a randomized controlled trial, it is very difficult to escape the risk of confounding. In this context, Verma’s[2] dismissive remark about having referenced “some statistical studies about confounding variables” is uncalled for.On a parting note, the failure to consider confounds in one's research can result in amusing results; an earlier article illustrates one such example.[6] A reader who does not understand research methodology will not understand research.