To the editor: The article entitled “Biomarkers trump behavior in mental illness diagnosis”1 seems to misinterpret the growing data on biomarkers in psychiatry in a dangerous way. Research on biomarkers is undoubtedly important, particularly for the insights it provides on pathophysiology and for its potential relevance in determining prognosis and response to treatment. But suggesting that biomarkers “trump behavior” in psychiatric diagnosis seems inappropriate, at least in the current state of medical knowledge. First, most studies dealing with biomarkers in psychiatric disorders, including the one mentioned in your article2, include patients diagnosed by current psychiatric classifications such as the DSM-IV, which are based on symptoms and clinical history. So, behavior is actually the gold standard in these studies, and stating that a biomarker is accurate based on these data merely means that it correlates well with symptoms. There is no way, therefore, that biomarkers could possibly trump behavior in such studies; on the contrary, if clinical symptoms are as dubious and misguiding for diagnosis as your piece seems to suggest, the markers that have been shown to correlate with them will also be so. Second, and perhaps more important, the assumption that psychiatric diagnoses as we know them today actually represent individual pathophysiologic entities is risky. As current definitions of psychiatric disorders are based on symptoms, the belief that a specific disorder corresponds to a specific neurobiological alteration is a mere assumption3, and one can easily make the case that one disorder may have many pathophysiologic bases or, conversely, that similar pathophysiologic processes can lead to different disorders in different people. Therefore, using biomarkers, genetic studies or any kind of objective biological data for psychiatric diagnosis would first require a remodeling of psychiatric nosology, which might lead to the spectrum of mental disorders being very different from what it is today. The points made above are not meant to invalidate the work on biomarkers, which can be valuable in many ways, such as in understanding the pathophysiology of mental disorders, predicting prognosis and response to treatment, and possibly helping early detection of illness in high-risk patients. Markers trumping behavior in psychiatric diagnosis, however, will only make sense when our understanding of the neurobiological bases of mental illness reaches a level that allows us to rewrite the science of psychiatry. This seems to be decades away, and one might argue that it may not happen at all. Until that day comes, advocating the use of markers in diagnosis is likely to end up doing more harm than good, leading clinicians more and more toward the easy solution of classifying people into disorders instead of listening to them and trying to understand their suffering.