Mental disorders take a staggering health and economic toll. The World Health Organization has estimated that unipolar depressive disorder alone is one of the leading causes of disability worldwide. Schizophrenia, bipolar disorder, autism and the many other psychiatric disorders only add to the misery. Yet progress in understanding the underlying causes of these conditions seems to be moving at a crawl. Genes are surely involved, but decades of futile hunting have made it painfully clear that the contribution of any single gene to disease is probably minuscule.

How to find these tenuous connections is a contentious scientific debate, with geneticists and neuroscientists at an apparent impasse (see page 154). These two communities must start working together more constructively if they are to crack this challenging problem and ensure that the millions of dollars now flooding into this field are not misspent.

Many geneticists believe that scanning the entire genomes of a massive number of patients will uncover weak gene candidates by sheer statistical power. Yet some neuroscientists dismiss these studies, questioning the utility of indiscriminately seeking a swath of genes that are weakly associated with a condition, and that are unlikely to have relevance to its biology. Another concern is that diagnoses of psychiatric disorders are often based on fuzzy, subjective and inconsistent criteria. The neuroscientists feel that such untargeted searches will be futile if the population labelled as having one condition — such as schizophrenia — in fact includes many different clinical disorders.

The antipathy is mutual. Some neuroscientists take what is to a geneticist an unacceptably small number of subjects — sometimes not even patients — and focus on establishing links between variants of a candidate gene and brain anatomy or function. Geneticists are often not persuaded by the statistics and question the focus on a few candidate genes whose link to disease has never been well established. Neither side seems to agree that the outcomes of the other's studies represent constructive leads. Nor is either side willing to compromise its standards. As a result, it is difficult to publish in the intersection between these two fields.

This conflict is unnecessary. Both communities will gain by learning from the other. Teams undertaking expensive genome-wide association studies should consult fully with clinical and research neuroscientists. If a single diagnostic label such as 'schizophrenia' is inadequate, then more detailed clinical information on patients must be collected when they are recruited. The studies might also benefit from more quantifiable measures of behaviour and neurobiology — such as laboratory assays of anxiety or the functioning of fear circuits in the brain.

And neuroscientists making early forays into genetic associations may find the hard lessons learned by geneticists over the years useful. Geneticists know about the statistical methods and criteria required to draw genetic conclusions that are persuasive and likely to stand the test of time.

The two sides should not lose sight of the fact that they share the same goal: to help patients. In the end, it is likely that both strategies and others will prove valuable in identifying risk factors and developing ways to counteract them. If genome-wide scans are appropriately designed and powered, they should yield the first clues. The neuroscientists' approach may then prove fruitful by showing what biological function is being performed by the genes identified, how that function leads to disease and how it can be altered.

When two fields such as these come together, lowering standards is not an option. Extracting psychiatric genes will require highly rigorous strategies, and on that geneticists and neuroscientists can agree.