Current classification systems (ICD-10 and DSM-IV) require a quantitative criterion for differentiating depressive states, suggesting a correlation between the number of symptoms, i.e., the pervasiveness of the syndrome, and the subtype of the illness. All the symptoms (within those contained in the diagnostic lists) are assumed to have comparable value. To investigate the relevance of the number and the type of symptoms reported by 196 patients suffering from depression, we compared the symptoms using independent indicators of severity such as the Clinical Global Index (CGI) and the social functioning subscale of the Global Assessment of Functioning (GAF). A second comparison using the same indicators was made between qualitatively distinct categories of DSM-IV and ICD-10 (i.e., melancholic v nonmelancholic, somatic v nonsomatic, and psychotic v nonpsychotic). There was evidence that increasing numbers of symptoms actually reflect higher levels of severity, but the categorizations that were mainly based on qualitative criteria (e.g., melancholia, somatic syndrome, etc.) usually attained better discrimination compared with those based on the number of symptoms. Moreover, certain symptoms (usually those indicated as endogenous) were more likely to be associated with greater severity and pervasiveness. Finally, the results clearly showed that different symptoms had different weight in establishing the gradient of severity.