<i>Background:</i> The present study investigated the interrater reliability of the diagnoses of schizophrenia (SCH), schizoaffective disorder (SAD), bipolar disorder (BPD) and unipolar depression (UPD) according to both DSM-IV and ICD-10, as well as the diagnostic congruence between the two classificatory systems. <i>Sampling and Methods:</i> Using the Composite International Diagnostic Interview, two trained psychiatrists simultaneously evaluated 100 inpatients and independently assessed the psychiatric diagnoses. The Cohen’s kappa coefficient was employed to estimate interrater reliability and diagnostic congruence between DSM-IV and ICD-10. <i>Results:</i> SCH was more frequent according to ICD-10 than DSM-IV criteria. Considering both diagnostic systems, all the four nosological categories, but ICD-10 SAD and DSM-IV UPD, were associated with interrater reliability coefficients above 0.50. The coefficient of the diagnostic congruence between DSM-IV and ICD-10 was inferior to 0.50 only for SAD. BPD was associated with the highest degrees of both interrater reliability and diagnostic congruence. <i>Conclusions:</i> The lack of an item excluding the occurrence of an affective syndrome among ICD-10 diagnostic criteria for SCH can account for: the larger frequency of SCH according to ICD-10 than DSM-IV; the unsatisfactory interrater reliability for the diagnosis of ICD-10 SAD, and the low diagnostic congruence for SAD.