One definition of a biomarker is ‘a naturally occurring molecule, gene, or characteristic by which a particular pathological or physiological process, disease, etc. can be identified.’ Over the past 10 years, there has been substantial investment in research looking for genetic biomarkers for psychiatric or cognitive phenotypes: NIH figures show grant awards on this topic rising from $145 million in 2009 to $562 million in 2018. The yield from this investment has, on the whole been disappointing. There have been advances in discoveries of rare mutations of large effect involved in the causation of conditions such as intellectual disability and autism, but for common psychiatric and developmental disorders and traits there have been mostly false leads and inconsistent findings. I will discuss the implications of this situation: in part, lack of progress may reflect methodological weaknesses that the field must overcome if we are to make progress in this area. However, the kinds of model of underlying process assumed by the biomarker agenda may be inappropriate for many common conditions, in which case we need to move to a different conceptualisation of gene-behaviour associations.