Due to the expense of collecting psychophysiological or similar measures, it is unlikely that large enough samples will be available to permit detecting variants that account for more than a trivial proportion of the variance (de Geus, 2010). Thus, what endophenotypes may offer in psychiatric genetics is less the promise of helping to identify novel genetic variants as helping scientists understand the neurocognitive characteristics of psychiatric disorders (Hall & Smoller, 2010). Alternative methods may be useful for increasing the sensitivity of genome-wide scans, such as Bayesian and network-based methods that make use of additional information or penalized regression approaches such as the lasso (Hastie et al., 2009), which permits selecting subsets of relevant SNPs. However, it may also be that different conceptualizations of the problem are required (de Geus, 2010; Hall & Smoller, 2010). For instance, trying to identify specific genetic influence on a phenotype, even an endophenotype, from a single cross-sectional snapshot may be somewhat akin to trying to understand the effects of gravity on a falling object while it is frozen in midair. Genes are expressed in particular
