The dichotomy between schizophrenia and manic-depressive illness is, as E. Kraepelin suspected, flawed; no unequivocal separation can be achieved. There are no categories of psychosis, but only continua of variation. However, the definition of nuclear symptoms by K. Schneider reveals the fundamental characteristics of the core syndrome--it is independent of the environment and constant in incidence across populations that have been separated for thousands of years. The associated genetic variation must be as old as Homo sapiens and represent a component of diversity that crosses the population as a whole. The fecundity disadvantage that accompanies the syndrome requires a balance in a substantial and universal advantage; this advantage, it is proposed, is the speciation characteristic of language; language and psychosis have a common evolutionary origin. Language, it is suggested, originated in a critical change on the sex chromosomes (the 'speciation event'--the genetic change that defined the species) occurring in East Africa between 100 and 250 thousand years ago that allowed the two hemispheres to develop with a degree of independence. Language can be understood as bi-hemispheric with one component function--a linear output sequence--confined to the dominant hemisphere--and a second--parallel distributed sampling occurring mainly in the non-dominant hemisphere. This mechanism provides an account of the generativity of language. The significance of nuclear symptoms is that these reflect a breakdown of bi-hemispheric coordination of language, perhaps specifically of the process of 'indexicalisation' (the distinction between 'I' and 'you') of self- versus other-generated references. Nuclear symptoms can be described as 'language at the end of its tether'; the phenomena and population characteristics of the nuclear syndrome of schizophrenia thus yield clues to the origin of the species.