The hypothesis of a psitive correlation between population density and levels of psycopathology was tested with an ecological study carried on in a low-income sector of the metropolitan region of Salvador - Bahia, Northeast Brazil. Prevalence rates of mental illness in children and adults, as well as scores of psychiatric simptomatology (collected by means of a cross-sectional survey with a representative sample of 1 549 adults and 848 children) were taken as indicators of the dependent variable. External and internal density, time of urban residence and average per capita income, measured directly from the same survey, were employed as independent variables. Multiple regression analysis showed that only income had a significant effect, but in the opposite direction, over the equations of all dependent variables. The increase in the R², which varied from 15 to 71%, was due mainly to variables not included in the hypothesis, such as average age, number of residents per dwellers and percent of migrants. The hypothesis of this study, in sum, are not confirmed by data, and such findings may be interpreted basically in terms of methodological problems of the study design.
