Using speech samples of 100 patients suffering from schizophrenia, bipolar illness and major depression, we addressed the question of the extent to which the linguistic abnormalities in the speech of these patients represent diagnosis-specific characteristics or constitute independent, syndrome-like dimensions of the illnesses. All speech samples were transcribed by a professional linguist who was blind to both identity and diagnosis of the patients. The majority of the deviant linguistic variables was found to be common to all three diagnostic groups under comparison, while only a few linguistic variables exhibited statistically significant between-group differences. On the other hand, when the respective variables were analysed as a multivariate entity, the variety of subtle between-group differences allowed us to discriminate between the diagnostic groups at an overall performance of 72.7% correctly classified patients. There was an almost complete lack of association between linguistic abnormalities and psychopathology syndromes. In particular, we found no correlation between the syndrome ‘formal thought disorder’ and the large variety of linguistic variables used in this investigation. In consequence, we conjecture that linguistically deviant speech characteristics represent an independent syndrome complex manifested at varying intensities across mental illnesses, and that this syndrome complex deserves greater attention, not only with respect to the principal understanding of the underlying disturbances, but also as a potential target of therapeutical intervention.

Evidence from previous studies has suggested that the inter-individual differences in human brain-wave patterns (EEG) are predominantly determined by genetic factors. In particular, the within-pair EEG concordance of monozygotic (mz) twins was found to be typically as high as r = 0.81 across channels and frequency bands, thus being comparable to that between repeated assessments on the same individual with typically r = 0.83. Yet our investigations into mz twins discordant and concordant for schizophrenia yielded a significantly reduced within-pair EEG concordance for both, the pairs discordant for schizophrenia and the pairs concordant for schizophrenia (with concordance for schizophrenia assessed through a syndrome-oriented approach). A multivariate discriminant function of EEG parameters distinguished in a reproducible way between affected and unaffected subjects at an overall performance of >75% correctly classified subjects, while the severity of illness, as derived from EEG-differences between affected and unaffected subjects, was closely related to the severity of illness as provided by psychopathology syndrome scores. Consequently, EEG anomalies associated with schizophrenia and manifested differently in the mz co-twins concordant for schizophrenia are likely the effect of nongenetic, pathologic processes that evolved independently in the co-twins' genetically identical brains once the illness began to progress. The existence of such nongenetic processes would suggest a modification of the standard phenotype-to-genotype search strategies of molecular-genetic studies that aim to link the schizophrenia phenotype to genetic vulnerability factors.