Schizophrenia has long been thought to be clinically heterogeneous. A range of studies suggests that this is due to genetic heterogeneity. Some clinical features, such as negative symptoms, are associated with a greater risk of illness in relatives. Affected sibling pairs are correlated for clinical and course features as well as subforms of illness, and twin studies suggest that this is due to genetic factors. This is further supported by findings that subjects from families linked to some chromosomal regions may differ clinically from those from unlinked families. Moreover, some genes may affect clinical features without altering susceptibility (ie are modifier genes). High-risk genotypes may have quantitative, rather than categorical effects, and may influence milder or subclinical phenotypes. Another recent finding is that nonpsychotic relatives may have personality features that resemble those of their affected relatives. These findings taken together suggest that there may be several classes of gene action in schizophrenia: some genes may influence susceptibility only, others may influence clinical features only, and still others may have a mixed effect. Furthermore, subsets of these classes may affect personality and other traits in nonpsychotic relatives. Understanding these classes of gene action may help guide the design of linkage and association studies that have increased power. We describe five classes of genes and their predictions of the outcomes of family, twin, and several types of linkage studies. We go on to explore how these predictions can in turn be used to aid in the design of linkage studies.