Everyone carries dangerous genetic mutations. But only in the past five years or so have researchers begun to use genomewide association studies (GWAS) to scour human genetic samples for the signals of individual variations. Typically, such studies assess hundreds of thousands of genetic variants in thousands of individuals sorted by traits: a certain height, perhaps, or asthma or obesity. Genetic variants that occur more frequently in one group than in another are subjected to stringent statistical analyses to determine whether associations between them and the traits are the result of biology or mere chance. As of 1 October, an online catalogue of GWAS contained nearly 700 publications linking some 3,000 variants to about 150 traits. The list of traits begins with abdominal aortic aneurysm and ends with YKL-40, a protein used as a biomarker for cancer. Other GWAS have identified correlations between genetic variants and smoking behaviour, sleep duration and general self-reported health. The catalogue is growing swiftly: 9 out of 16 research articles in the October issue of Nature Genetics report GWAS. In their current incarnation, GWAS are running into a problem of diminishing returns. By collecting ever-larger samples, researchers are able to find more and more variant–trait associations, but these tend to have smaller and smaller effects. In fact, small effect sizes have been a hallmark of GWAS ever since the studies began. Originally, researchers hoped to find associations in which people carrying one variant would be several times more likely to have a trait than those carrying another. Instead, the effects found have been much more modest. An analysis published in June 2010 (ref. 1) pooled findings from several published GWAS that had each associated given traits with single nucleotide polymorphisms (SNPs) — the simplest and most common type of genetic variant, in which one DNA letter is changed to another. Extrapolating from previous findings, the researchers calculated that 201 SNPs associated with height could explain about 16% of genetic variance, 142 SNPs associated with Crohn’s disease could explain about 20%, and 67 SNPs could explain about 17% of genetic variance in each of three common cancers. Although genetic variants with small effects can still help to uncover fundamental biology with therapeutic implications, researchers hunting for those with larger effects are pinning their hopes on several advances: an onslaught of newly discovered simple polymorphisms, the ability to assess more complicated variants (see ‘The tough new variants’) and multiplying applications of sequencing. If the human genome were an archaeological site, these options would be equivalent to canvassing continents with metal detectors of varying convenience and reliability, or picking a handful of sites for a full excavation.