Heritability estimates derived from twin and adoption studies are used in psychiatry and the humanities as evidence for the influence on phenotypic traits or diseases and as a measure of the extent of this influence. Despite their popularity, these estimates have a number of fatal conceptual and empirical flaws, making them largely useless for human research. Heritability is defined as a measure of the proportion of total phenotypic variance in a trait that is accounted for by variance. It can take on values between 0 and 1, with 1 indicating that 100% of the total phenotypic variance is accounted for by variance. If, for example, all the differences in hair colour in a human population can be accounted for by differences between the individuals, hair colour is said to be 100% heritable. While the heritability statistic as defined only addresses vs environmental contributions to the difference in phenotypic traits, it is often interpreted as evidence for vs environmental contributions to the phenotypic traits themselves. This may seem like an innocuous restatement of the same fact, but it is a crucial and fallacious shift in interpretation, because what causally accounts for the difference between two things X and Y is often just a minimal subset of what accounts for X (or Y) itself. For example, a single switch may cause the difference in destination between two trains that start from the same city, but what accounts for the fact that a train reaches any destination at all involves a much richer causal story including trains and how they are built, railway networks and the desire of human beings to get from one location to another by means of public transportation. Psychiatry, and the study of human development in general, is primarily interested in traits, not trait differences. The traits of psychiatry are psychiatric disorders, and the most issue in psychiatry concerns the understanding of the causal mechanisms leading to a phenotype, in the service of finding therapeutic interventions. Heritability estimates, however, are useless for the search for effective treatment, because by themselves they give no indication of where in the genome to look for as potential targets for therapeutic intervention. Not even the general indication that genes are important for a trait or is helpful, since it is a fundamental fact of developmental biology that both genes and non-genetic factors (such as the environment) are always involved in the causation of any trait or phenotypic feature. Neither genes nor the environment work in isolation. In the process of organismic development and non-genetic resources interact in many places in a multitude of ways, and the outcome always reflects the joint interaction of the various resources. Being told that the heritability of, say, schizophrenia is 90% does not tell one anything about whether and how much genes are involved in schizophrenia (they are always involved) nor where in the causal hierarchy of the organism therapeutic intervention might be most effectively aimed at. In particular, a high heritability does not mean that treatment on the level will be most effective. This is forcefully illustrated by the case of the genetic disease phenylketonuria (PKU), where a gene defect is seen as the basis of the disorder, but treatment most effectively operates through dietary (i.e. environmental) manipulation. The case of PKU also illustrates another fact: that involvement in a trait or does not mean that the trait/disease is difficult or impossible to change. Although a common idea, it is entirely fallacious. The question of how easily a phenotypic trait can be changed is a purely practical one and has nothing whatsoever to do with the question of whether the trait has factors among its causes. What matters most for successful treatment efforts is as complete as possible an understanding of the causal processes leading to the trait or disease. Heritability estimates are of no help in this endeavour.