We are delighted that our article has stimulated discussion about strategies for gene discovery in schizophrenia. We agree that schizophrenia, like other complex traits, will be influenced by a large number of genetic and epigenetic events with a spectrum of effects. Both rare alleles of large effect and common alleles of modest effect are likely to be discovered (Reference Craddock, O'Donovan and OwenCraddock et al, 2007). Rare severe-effect alleles are fully compatible with familial patterns of schizophrenia because many (perhaps most) such alleles have arisen de novo in the present or recent generations. De novo mutations play havoc with predictions of conventional recurrence risk models. For example, de novo meiotic mutations (in the parental germline) increase disease concordance among monozygotic but not dizygotic twins. In contrast, de novo mitotic mutations or epigenetic events (in early embryogenesis) reduce concordance among both monozygotic and dizygotic twins.

Genetic association studies are not the most straightforward path to gene discovery for schizophrenia. Individually rare alleles cannot be identified by comparing frequencies of common alleles among unrelated patients with controls, even with enormous numbers of well-diagnosed patients, properly matched controls and very dense (and expensive) screening tools. To the extent that rare alleles are important to schizophrenia, study designs based on a naive ‘common disease–common allele’ model will yield variable and non-replicable results (Reference King, Ahsan, Susser, Susser, Schwartz and MorebiaKing et al, 2006).

Characteristic patterns of age at onset, gender differences and brain changes associated with schizophrenia are fully compatible with causal influences of rare severe-effect events, either genetic or epigenetic. Each such event alters the expression, timing or function of one of a very large number of genes. The products of these genes converge in common pathways. Aberrations of a pathway by any of multiple mechanisms may lead to clinically similar disorders.

Crow's proposition that schizophrenia arises from the disruption of uniquely human genetic elements is very appealing. This premise, however, need not narrow the search for causes, genetic, epigenetic or environmental. Human speciation likely occurred primarily as a result of regulatory changes in genes, rather than common polymorphisms leading to changes in gene sequence (Reference King and WilsonKing & Wilson, 1975). The extraordinary number of repeated elements in the human genome gave rise to a vast number of new genes and regulatory mechanisms. Their architecture also created an increased risk for copying errors. Thus, one cost of the genomic complexity that enabled human brain development may be a de novo error rate that results in the maintenance of schizophrenia in the population.

Autism has recently been shown to be associated with a significantly increased frequency of rare de novo mutations (Reference Sebat, Lakshmi and MalhotraSebat et al, 2007). These results presage the identification of many more rare mutations associated with other neurodevelopmental illnesses, as advances in technology enable the detection of ever-smaller genomic lesions. The ultimate resolution of this debate lies in gene discovery, for which we encourage the application of study designs most likely to be fruitful.