In contrast to the hypothesis-free linkage approach, candidate gene studies focused on genes, and variants within those genes, that were hypothesized to have biological relevance to the outcome of interest. In this way, candidate gene studies had the advantage of being more precise than linkage studies in that they had the potential to pinpoint specific genes or genetic variants, rather than just specific chromosomal regions. However, they relied on the investigator to correctly “guess” what genes were biologically relevant to the outcome. Despite thousands of candidate gene publications on behavioral phenotypes, the approach remains controversial, and very few candidate gene findings are widely accepted within the genetics community. Over time, it has become clear that the genetic architecture of behavioral traits is highly complex, that effect sizes of genetic polymorphisms are likely to be small (see below), and that scientists’ ability to predict a priori which genes are likely to be relevant to a behavioral outcome has been very poor (F.J. Bosker et al., 2011; Colhoun, McKeigue, & Davey Smith, 2003; Need et al., 2009; P.F. Sullivan et al., 2008).
